WO1995001053A1 - Processeur d'images - Google Patents

Processeur d'images Download PDF

Info

Publication number
WO1995001053A1
WO1995001053A1 PCT/JP1994/001029 JP9401029W WO9501053A1 WO 1995001053 A1 WO1995001053 A1 WO 1995001053A1 JP 9401029 W JP9401029 W JP 9401029W WO 9501053 A1 WO9501053 A1 WO 9501053A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
memory
writing
reading
processing
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/JP1994/001029
Other languages
English (en)
French (fr)
Inventor
Tatsuhiko Demura
Kazukuni Kitagaki
Goichi Ootomo
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Toshiba Corp
Original Assignee
Toshiba Corp
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Toshiba Corp filed Critical Toshiba Corp
Priority to DE69425720T priority Critical patent/DE69425720T2/de
Priority to KR1019950700678A priority patent/KR100252838B1/ko
Priority to EP94918576A priority patent/EP0660608B1/en
Publication of WO1995001053A1 publication Critical patent/WO1995001053A1/ja
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/42Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
    • H04N19/43Hardware specially adapted for motion estimation or compensation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/503Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
    • H04N19/51Motion estimation or motion compensation
    • H04N19/523Motion estimation or motion compensation with sub-pixel accuracy
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/42Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
    • H04N19/423Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation characterised by memory arrangements
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/503Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal prediction
    • H04N19/51Motion estimation or motion compensation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04NPICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/60Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/61Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding

Definitions

  • the present invention relates to an image processing device using motion compensation predictive coding, and an image processing device for decoding a code-compressed image.
  • the basic technology in this technology is image compression / decompression technology.
  • MPEG1 Motion Picture Expert Group part 1
  • MEPEG 2 part 2 of IS0 (International Standard Organization)
  • the MPEG 1 standard is a transmission standard for relatively low bit rate non-interlaced images
  • the MPEG 2 standard is a relatively high bit rate non-interlaced image. It is a transmission standard for source or interlaced images.
  • Inter-frame predictive coding is coding that uses temporal correlation of images.
  • the transmitting side transmits only the difference between the previous screen and the current screen, and the receiving side stores the difference between the previous screen and the transmitted screen and adds the transmitted difference value to the previous screen. Reconstruct.
  • the processing is expressed as (1)
  • the transmission bit width can be reduced for images with little change or little movement (images with high temporal correlation). Images with moving power, motion, and motion can be added to the bit width. Therefore, for images with dynamic power, the bit width can be reduced by adding processing called motion compensation.
  • the transmitting side evaluates the amount of movement between the previous screen and the current screen, and transmits the amount (movement vector) together with the difference value.
  • the value of the current screen is reconstructed by adding the difference value to the value of the position that has moved by the motion vector from the previous screen and adding the difference value. The processing is shown in equation (2).
  • D (X, y, t) D (x + mvx, y + mvx, t -x) + ⁇ D ( ⁇ , y, t) ... (2) where mvx and mvy are each The motion vectors are in the x and y directions.
  • Equation (3) shows the interpolation processing of the four-point force.
  • [X] represents the largest integer that does not exceed X with respect to the rational number X.
  • restoration processing As described above, in the image mediation apparatus using the motion compensation prediction coding, the interpolation of the reference image and the addition of the difference value (hereinafter referred to as restoration processing). ) is necessary .
  • the structure of the transmitted image includes a frame structure and a field structure.
  • scanning lines are sequentially transmitted for each line as shown in F1G.1 ⁇ (a).
  • field structure interlaced scanning is performed as shown in Fig. 10 (b), and in the first field, the even lines are used, and in the second field, the even lines are used. Transmits odd lines.
  • frame prediction and field prediction in the method of calculating the motion vector.
  • frame prediction the reference image is treated as a frame structure
  • field prediction the reference image is treated as a field structure (when the reference image is A frame-structured platform is treated as an image of even or odd lines only.
  • the MPEG 1 standard only supports the frame structure.
  • An example of a decompression circuit for motion compensation predictive coding is shown in F1G.11.
  • adder 2 sets the reference image This is a block in which the interpolation process is performed, and the adder 3 performs a restoration process.
  • the reference image memory is a frame memory.
  • the corresponding reference image is accessed as needed.
  • the read operation is performed at the same time as the read operation
  • the restoration process is performed at the same time as the read operation.
  • the interpolation process and the restoration process are performed at the same time.
  • processing such as inverse quantization and inverse cosine conversion is performed by the arithmetic unit 6, which causes an equivalent delay. Therefore, it is necessary to adjust the timing for accessing the reference image according to the output of the result of those processes.
  • An image transmitted by a signal of the MPEG 2 standard can take both a frame and a field as well as an image structure and a prediction method of a transmitted difference.
  • the restoration processing is performed.
  • a frame / field conversion device must be set up before this, and the structure of the difference image and the prediction method must be combined.
  • An example of a decompression circuit for motion compensation predictive coding in that case is shown in F1G.13.
  • the adder 2 is a block for performing the interpolation processing of the reference image
  • the adder 3 is a block for performing the restoration processing.
  • frame / no-field conversion is performed in the intermediate memory.
  • the reference image memory is a frame memory. Divide the entire frame into smaller screens (for example, 8x8 blocks) The contents of the processing when the processing is performed by using the timing chart of FIG.
  • the reference image memory 5 the small image and its peripheral pixels are read out, interpolated by the adder 2, and the result is written to the intermediate memory 4.
  • the difference data is subjected to processing such as inverse quantization and inverse cosine conversion in the arithmetic unit 6. These processes are performed on the small screens that are simultaneously performed during the period (1) of F1G.14, and when the processes are completed, the intermediate process is continued.
  • the reference image is read using the same frame field structure as the difference data, and is added to the difference data input from outside by the adder 3. Then, restoration processing is performed. These processes are performed simultaneously during the period (2) in FIG.
  • an image stretching device that does not have intermediate memory, reads a reference image in accordance with the processing of difference data, and performs interpolation and restoration processing is used for interpolation.
  • an adder for restoration and an adder for restoration processing are necessary, and the hardware amount is large.
  • a motion compensated predictive coded image decompression device that has intermediate memory and performs frame-no-field conversion processing requires a hardware If the amount becomes large, there is no over-lapping during the time when the adder operates before and after the intermediate memory, and the other side performs processing. While there are times when the other is not in use, there is a drawback to saying that this is a very inefficient system.
  • code input buffer for synchronization, reference image for motion compensation, frame memory, display image for frame / field conversion Three types of memory devices, frame memory, are required.
  • these memories cannot be integrated in the same LSI as the decoding device at the current LSI density. Therefore, it must be set outside the LSI as external memory. If different memory is allocated to each application, the number of wirings, circuit scale, and amount of memory will increase.
  • One memory as the external memory Many methods are used to divide the image, display image, and code input buffer and share them.
  • the input code is written to the code input buffer, and the code input buffer is read.
  • Read the reference image for motion compensation reconstruct the reference screen after decoding, write to the reference image, and display the decoded image Readout of the displayed image for reading, the following five types of memory access occur.However, the input code that allows data to flow without interruption is input to the code input buffer.
  • writing and reading images that must always be displayed reading out the displayed images cannot overflow or disappear, and the processing cannot be overridden. Because of the head, it is difficult to stop reading and writing the reference image once it has been performed. Yes.
  • the first is to send an access request to the bus arbitration device each time memory access is required, and the bus arbitration device sends a request signal generated from each block.
  • This is a priority method in which arbitration is performed according to priority.
  • As another method there is a polling method in which requests are processed in order without priority.
  • the priority is given when the memory access amount increases. Memory access with a lower rank is not performed, and the waiting state increases.
  • One of the processing The number of heads increases, the decoding power of one frame image increases, and it becomes impossible to perform it during one frame display time, and the input code and the output image are synchronized. There was a problem of going away. In the case of the latter method, continuous memory access cannot be performed. Therefore, when DRAM is used for external memory, page switching is performed. One head force increases, and the memory band width becomes insufficient.
  • the motion compensation will be applied to the part with a relatively large amount of redundancy in the spatial direction (moving image). Since it is hard to apply, DCT + quantization and DPCM are used to reduce the redundancy of the space axis method. In this case, the DPCM only covers the DC component that has the highest correlation with the adjacent block.
  • the motion compensation is applied to the part with relatively much redundancy in the time axis direction (images that are not moving so much), and only the motion vector and the error component are data.
  • DCT + quantization is performed for each block of a certain size for the error component.
  • DPCM is not used.
  • FIG. 15 shows a conventional image processing apparatus for restoring data encoded as described above into an image.
  • a variable-length code is decoded by a code reading circuit 30 to obtain a flag and a DCT coefficient.
  • the DCT coefficient of the difference data is IDCT performed by the inverse quantization (IQ) / inverse discrete cosine transform (IDCT) circuit 31.
  • IQ inverse quantization
  • IDCT inverse discrete cosine transform
  • the difference data is obtained, and simultaneously, the FI
  • the motion vector is reconstructed by an algorithm such as that shown in G.16, and the motion vector is reconstructed by the motion compensation circuit 33 using the motion vector.
  • the forward, backward, or both image data is extracted from the image memory 34, and is added to the difference data overnight by an adding circuit 35 to generate restored pixel data. .
  • a conventional image decompression device that does not have intermediate memory, reads a reference image in accordance with the processing of difference data, and performs interpolation and restoration processing.
  • Large-scale circuit, moving-compensated predictive coded image expansion with intermediate memory and frame-no-field conversion In the image decompression device that performs the above, there is a problem that the circuit standard of the adder becomes large, and the power and the time when the adder is not used are present. .
  • the conventional general bus arbitration function requires a large amount of overhead when accessing external memory. There was a weakness.
  • the DPCM processing of the DC component of the T coefficient uses the decoded flags to perform conditional judgments and various operations as shown in FIG. 16 and FIG. 17 using the decoded flags.
  • a sequencer is required, which can be a large hardware. Compression of image data for which various standards and applications are conceivable ⁇ Since decompression techniques may differ in the way of interpreting and calculating flags For the above sequencer, the one that can be controlled by software has better adaptability and expandability, but requires more memory such as instruction memory. The amount of dwell increases.
  • the conventional image processing device requires two sequencers and therefore has a tendency to increase the amount of hardware. Disclosure of invention
  • the present invention has been made in view of this point, and an object of the present invention is to provide an image processing apparatus which operates with a small circuit scale and performs compensated predictive coding image expansion. There are. Another object of the present invention is to use an internal buffer for reading and writing of external memory and to use the external buffer for specific purposes. It is an object of the present invention to provide an image processing device capable of efficiently accessing memory. Further, another object of the present invention is to provide an image processing device with good hardware use efficiency.
  • an image processing apparatus Is composed of an image memory and an adder for the processing area for which motion compensation is performed, and the same adder is time-multiplexed for interpolation and restoration without time overlap. And use it for processing.
  • decoding means for decoding an input code and restoring an image, and receiving a code string input from the outside and writing the code string to an external memory.
  • Input code writing means for writing, input code reading means for reading a code string stored in external memory, and transferring the code string to the decoding means;
  • Image reconstructing means for reading out a reference image stored in memory and creating a display image by adding the reference image and the image restored by the decoding means;
  • Image writing means for writing the image created by the image reconstructing means into external memory, and writing by the image writing means for displaying an image externally Reading the loaded image from external memory Reading the displayed image Means for connecting the external memory to the image reconstructing means, the image writing means, the input code writing means, the input code reading means, and the display image reading means.
  • a memory arbitration means for selecting the connection means from among the connection means.
  • the memory bus arbitration means writes the above-mentioned input code to the code input buffer, reads out the code input buffer power, reads out them, and compensates for motion.
  • the reference image is read out at the same time, the reference image is reconstructed after decoding, the reference image is written, and the decoded image is displayed. The image is read out and the
  • the first group for reading out images and displaying the decoded images are the same as the first group for reading out images.
  • the reference image is read out at the same time as compensation, and the reference image is reconstructed after decoding.This is divided into the second group of writing to the reference image.
  • access rights were granted only for a certain period by the polling method, and all requests for the first group were completed.
  • the right to access is transferred to the second group, and when the right to access the second group is transferred, the second If there is an access request in the group, the memory access is not performed, and the memory access in the second group is accessed once. Granting access rights will cause other access requests to be rejected until the end of the series of memory accesses.
  • discrete cosine transform and motion compensation are used, and only a part of the discrete cosine transform coefficient where no motion compensation is performed is used.
  • the image data is compressed by DPCM conversion.
  • the DPC of the discrete cosine transform coefficient is used in the device that restores the image data from the code of the encoding method. It has a program-controllable sequencer that can perform both the processing of M and the reconstruction of the motion vector.
  • the circuit size of the adder of the image processing device for performing motion compensation predictive coding image expansion is reduced to approximately 12. Power ⁇ I can do it.
  • the buffer for writing the code input buffer the buffer for reading the code input buffer, and the display image
  • each memory access is requested from the memory arbitration device, and the reference image is displayed.
  • the processing involved in the internal decoding reads out the reference image, and when the reference image needs to be written.
  • the memory bus arbitration device can write the code input buffer, read the code input buffer, and read the display image. Process requests in a certain order. The writing of the reference image and the reading of the reference image are performed by writing the code input buffer, reading the code input buffer, and reading the display image. Perform the processing when the access request power is no longer available.
  • the reference image is read and written continuously. Since the reference image can be read out and written, the processing that should be performed after writing can be performed promptly. With fewer heads, a larger image can be obtained. be able to . Also, when compared to the priority scheme, reduce the number of page switches associated with accessing the DRAM without increasing the internal buffer. It allows you to handle larger images without having to use expensive high-speed memory.
  • the image processing apparatus based on the other embodiments described above processes various flags after decoding by a single software-controllable sequencer so as to perform IQ processing.
  • IDCTI IDCTI
  • FIG. 1 is a block diagram of the image processing apparatus according to the first embodiment.
  • FIG. 2 is a configuration diagram of the image processing apparatus according to the first modification of the first embodiment. A timing chart showing the processing contents of the image processing device.
  • F 1 G.3 is a timing chart showing the processing contents of the image processing apparatus according to the second modification of the first embodiment.
  • FIG. 4 (a) to FIG. 4 (d) are diagrams illustrating an image structure in the process of a restoration process in the first embodiment.
  • FIG. 5 is a diagram showing an example of an internal file according to the first embodiment.
  • FIG. 6 is a block diagram showing the configuration of the image processing apparatus according to the second embodiment.
  • FIG. 7 is a configuration diagram of the image processing apparatus according to the third embodiment (F1G.8 is an example of a program of the image processing apparatus according to the third embodiment).
  • FIGS. 9 (a) and F1G.9 (b) are diagrams illustrating the accuracy of the motion vector according to the first embodiment.
  • FIG. 10 (a) and FIG. 10 (b) are diagrams illustrating a frame structure and a field structure of an image according to the first embodiment.
  • FIG. 11 is a block diagram of a conventional image processing apparatus without frame / final conversion.
  • F1G.12 is a timing chart showing the processing contents of a conventional image processing apparatus without frame-to-frame conversion.
  • FIG. 13 is a block diagram of a conventional image processing device with frame-Z field conversion.
  • FIG. 14 is a timing chart showing the processing contents of a conventional image processing apparatus having a frame / field conversion.
  • FIG. 15 is a block diagram of a conventional image processing apparatus.
  • FIG. 16 is an explanatory diagram of the algorithm for motion vector reconstruction.
  • FIG. 17 is an explanatory diagram of an algorithm for DPCM processing of the DC component of the DCT coefficient.
  • FIG. 18 is a diagram that explains the meaning of each command shown in FIG. 18 and FIG. 17.
  • FIG. 1 is an adder that performs interpolation and restoration processing
  • 4 is an intermediate memo that holds a small screen (for example, an 8 ⁇ 8 block).
  • Reference numeral 5 denotes a frame memory which can hold a reference image for one frame.
  • the timing chart of the processing of the first modified example is shown in F1G.2.
  • processing such as inverse cosine conversion is performed by the six computing units. These processes are performed simultaneously during the period (1) of F1G.2.
  • the reference image is read from the intermediate memory in step 4 and added to the difference data input from the outside by the same 1 adder as that used in the interpolation processing.
  • restoration processing is performed. These processes are performed simultaneously during the period (2) in FIG.
  • the reference image interpolation processing is performed during processing such as inverse cosine conversion of difference data, and the same adder is subjected to interpolation processing and restoration processing. Therefore, the circuit size of the adder can be reduced to almost 1/2, and a lean and efficient system can be realized.
  • FIG. 3 shows a timing chart of the processing of the second modified example.
  • the frame Z field conversion is performed using the intermediate memory of 4.
  • the processing contents of the base unit that divides the entire frame into small screens (for example, 8 ⁇ 8 blocks) and processes them are shown in FIG. This will be explained using a mining chart.
  • the difference data is subjected to processing, such as inverse cosine conversion, by the six computing units. These processes are performed simultaneously during the period (1) in FIG.
  • FIG. 4 shows the structure of the reference image at the time of field prediction.
  • a 1 to a 20 correspond to pixels.
  • Pixels connected by solid lines correspond to the first field, and pixels connected by dotted lines correspond to the second field.
  • the data written in the field unit is read out in the frame unit. That is, bab 2, b 3, b 4, b 9, bl ⁇ , b 11, b 12, b 5, b 6, ⁇ ⁇ ⁇ ⁇ Difference data cl, c 2, c 3, c 4, c 5, c 6, c 7, c transmitted in the frame structure 8, c 9, c 10, ⁇ (FIG. 4 (c)) Calculated and restored dl, d 2, d 3, d 4, d 5, d 6, d 7, d 8, d 9, d 10, ⁇ ⁇ ⁇ (FIG. 4 (d))
  • the internal processing of the reference image is performed during processing such as inverse cosine conversion of the difference data, and the same adder is subjected to internal processing and restoration processing. Therefore, the circuit size of the adder can be reduced, and the conversion of the frame Z field is also performed in the intermediate memory. Also, there is no additional memory, and the circuit size can be reduced to almost one-two of the conventional example when the prediction method is a prediction in both forward and backward directions.
  • FIG. 6 is a block diagram showing a configuration relating to the image processing apparatus of the second embodiment.
  • this decoding device 101 includes a decoding processing unit 102, an input code writing unit 103, an input code reading and reading unit 104, and a display image.
  • a memory device 114 is connected to the outside of the decoding device 101.
  • An input code from the outside is input to an input code writing section 103 and stored in a buffer 109.
  • a write request is sent to the media bus arbitration unit 108 and the memory bus arbitration unit is sent.
  • the memory access permission signal is returned, the memory device 113 and the address generator 112 cause the memory device 114 to return. Writes a fixed amount of data in the buffer 109.
  • the code reading unit When the accumulated amount of the buffer 110 falls below a certain amount, the code reading unit reads out the data to the memory bus arbitration unit 108 and sends a read request to the memory bus arbitration unit 108. When the memory access permission signal is returned from the arbitration unit 108, the memory is accessed by the memory bus 113 and the address generation unit 112. The device 114 reads the data in a predetermined amount and stores it in the buffer 110. The code reading unit sequentially inputs the data stored in the buffer 11 ⁇ to the decoding unit 1 ⁇ 2.
  • the reference image reconstruction unit 106 reads out and sends a request to the memory arbitration unit 108, and When the memory arbitration unit 108 returns a memory access permission signal, the memory arbitration unit 108 and the address generation unit 112 return to the memory.
  • Required data from the device 1 1 4 The entire image is read, and the image that is associated with the prediction is reconstructed using the read data and the data decoded by the decode processing unit.
  • the reference image writing unit 1 ⁇ 7 sends a write request to the memory bus arbitration unit 108 when the image data is reconstructed by the reference image reconstructing unit.
  • the memory access arbitration unit 108 returns a memory access enable signal
  • the memory bus arbitration unit 108 and the address generation unit 112 cause the memory access arbitration unit 108 to return. Write the reconstructed image data to the memory device 1 14
  • the display image reading unit 105 reads out the data stored in the buffer 111, performs necessary processing, and outputs the processed data to the outside.
  • the memory arbitration unit 1-8 sends an s request to the memory arbitration unit 1-8 and sends a memory request.
  • the memory access permission signal is returned from the memory arbitration unit 108, the memory and the address generation unit 112 enable the memory device 114 Then, the data is read only in a predetermined amount and stored in the buffer 11.
  • the memory arbitration unit has a fixed memory access request of the display image reading unit 105, input code writing unit 103, and input code reading unit 1-4. Request processing in the order shown.
  • Writing of the reference image Reading of the reference image is performed by the memory of the display image reading section 105, the input code writing section 1 ⁇ 3, and the input code reading section 104. Generates a memory access permission signal when there is no longer a service request.
  • the display image reading unit 105 The data of the input code writing section 103 and the input code reading section 104 buffer 9-1 to 11-1 become empty or full.
  • the data could not be interrupted and the processing would not be interrupted.
  • the reference image could be written and the reference image could be read continuously.
  • the overhead associated with decoding and memory access can be reduced.
  • F 1 G.7 shows an embodiment of the image processing apparatus of the third embodiment.
  • This image processing device uses a single software-controllable sequencer to process the various flags after decoding, and to apply the DCT coefficient to the IQ * IDCT circuit and the motion compensation circuit. This is a configuration that gives a motion vector.
  • the code decoding circuit 10 does not decode the variable-length code and finds the flags and DCT coefficients.
  • the DCT coefficient of the differential data is subjected to IDCT in the IQ IDCT circuit 11 to obtain the differential data, and the data is simultaneously processed.
  • the motion vector is reconstructed by the sequencer 12 controllable by the program, and the motion compensation circuit 13 is used to reconstruct the image memory by using the motion vector.
  • the forward and / or backward image data is extracted from 14 and added to the difference data by an adder circuit 15 to create restored pixel data.
  • the sequencer 12 used in the image processing apparatus according to the third embodiment can execute a program as shown in F1G.8 and can operate the program. It is possible to perform time-division reconfiguration of files and DPCM processing of DC components of DCT coefficients.
  • the circuit scale of the adder of the image processing device for performing the motion compensation predictive coding image extension is reduced to almost 1 Z2. This makes it possible to construct a lean and efficient system.
  • the internal buffer for reading and writing the external memory is separately provided, and the reading of the reference image and the Z writing are performed. Is performed continuously. As a result, it is possible to efficiently access the memory without increasing the overhead associated with accessing the external memory. I can do it.
  • the window can be suppressed to a necessary minimum. Furthermore, it is possible to respond to various standards and applications only by developing software without increasing hardware, and it is possible to deal with various standards and applications. It is possible to provide an image processing device with high use efficiency.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Color Television Systems (AREA)

Description

明細書 画 像処理装 置 技術分野
本発明 は 動 き 補償予測符号 化 を用 い た 画 像処理装置 、 及 び符号圧縮 さ れ た 画 像 を復号す る た め の 画 像処理装置 に 関す る 。 背景技術
近年、 情報処理機器 の発展 は め ざ ま し く テ レ ビ電話 テ レ ビ会議 の 装置 も 数多 く 提案 さ れて い る テ レ ビ会議 テ レ ビ電話 な ど の 画 像伝送 の 分野 に 於 い て 基盤技術 と な る の が画像圧縮伸長の 技術で あ る 。 現在 画 像圧縮 の ffl際標準ィ匕作業力く I S 0 ( International Standard Org an i zat i on ) の M P E G 1 ( Motion Picture Expert G r oup part 1) , M E P E G 2 ( 同 part 2) で行わ れて い る 。 M P E G 1 規格 は比較的低 ビ ッ ト レ ー ト の ノ ン イ ン タ ー レ ー ス 画 像 の 伝送規格で、 M P E G 2 は比較 的高 ビ ッ ト レ ー ト の ノ ン ィ ン タ 一 レ ー ス あ る い は ィ ン タ ー レ 一 ス 画像 の 伝送規格で あ る 。
画 像圧縮伸 長 の 方式 に は さ ま ざ ま な 方式が提案 さ れ て い る が、 そ の 一つ に 動 き 補償予測符号 化方式が あ る 。 ま ず、 フ レ ー ム 間予測符号 化方式 に つ い て 説明 す る 。 フ レ 一ム 間予測符号 化 は 画像 の 時間 的相関 を利用 し た 符号 化 で 、 送信側で は前画面 と 現画面差分 の み を 伝送 し 、 受信 側で は 蓄 え て お い て前画面 に 、 伝送 さ れ た 差分 値 を加算 す る こ と よ り 、 現画 像 を再構成す る 。 そ の 処理 を ( 1 ) 式 不 o
D(x.y,t)=D ( x,y.t-l)+ Δ D(x , y , t) … (i)
こ の 方式で は変 化、 あ る い は 動 き の 少 な い 画 像 (時 間 的 に 相関が高 い 画 像 ) に つ い て は 伝送 ビ ッ ト 幅が削減で き る 。 し 力、 し 、 動 き が あ る 画 像で は む し ろ ビ ッ 卜 幅力く增 加 し て し ま う 。 そ の た め 、 動 き 力《 あ る 画 像 に 対 し て は 、 動 き 補償 と い う 処理 を追加す る こ と に よ り 、 ビ ッ ト 幅 を 削減す る 。 動 き 補償処理で は 、 送 信側で前画面 と 現画面 の 動 き 量 を 評価 し 、 そ の量 (動 き べ ク ト ル ) を 差分 値 と 共 に 伝送す る 。 受信側で は前画面 か ら 動 き べ ク ト ル分 だ け遷移 し た 位置 の 値 に 差分 値 を加 え て現画面 の 値 を 再構 成す る 。 そ の 処理を ( 2 ) 式 に 示す。
D ( X , y , t ) =D ( x + mvx, y + mvx , t -丄) + Δ D ( χ , y , t ) ··· (2) こ こ で、 m v x , m v y は そ れ ぞれ x 方 向 , y 方 向 の 動 き べ ク ト ルで あ る 。
動 き べ ク 卜 ノレ の 取 り 方 に は い く つ か の 方式力く あ る 。 動 き べ ク ト ル の 精度が画素単位の 場 台 は動 き べ ク ト ル の 指 し 示す位置 の 画素 の 値 D ( x + m v x , y + m v y , t 一 1 ) は存在す る ( F! G . 9 ( a ) 参照) 。 し 力、 し 、 動 き べ ク ト ル の 精度力《 1 Z 2 単位の 場合 は動 き べ ク ト ル の 指 し 示す 位置 に 画素が存在 し な い 場合が あ る ( F ί G . 9 ( b ) 参照) 。 そ の 場合 に は近傍 の 2 点 も し く は 4 点か ら 内揷 し て そ の 点 の 値 を 得 る ( F 1 G . 9 ( b ) 参照) 。 4 点 力、 ら の 内 挿処理 を ( 3 ) 式 に 示 す。
D(x - y. l) - !l)([x*mvx] . [y + mvy] . t-l )+l)([x+mvx] + | . [y + nivy] . (-| )
+ D([x+mvx] . [y + nivy]+ l . t-| ) +|)([x*mvx] . [y*mvy] t | .1-1 )1 /4 +厶 D(x . y, t) …… (3) こ こ で 、 [ X ] は有理数 X に 対 し て X を 越 え な い 最大 の 整数 を 表す。
以上 の よ う に 、 動 き 補 償 予測符号 化 に よ る 画 像 仲 長 装 置 に 於 い て は 、 参照画 像 の 内 挿処理 と 差分 値 と の 加 算 ( 以 後 復元処理 と い う ) が必要で あ る 。
伝送 さ れ る 画 像 の 構造 に は フ レ 一 厶 構造 と フ ィ ー ル ド 構造が あ る 。 フ レ ー ム 構造 で は F 1 G . 1 ◦ ( a ) の よ う に 走査線 を 1 ラ イ ン 毎 に 順次伝送す る 。 フ ィ ー ル ド 構造 で は 、 Fig 1 0 ( b ) の よ う に 飛 び越 し 走査 を 行 い 、 第 1 フ ィ 一 ル ド で は 偶数 ラ イ ン を 、 第 2 フ ィ ー ル ド で は 奇 数 ラ イ ン を 伝送す る 。
ま た 、 動 き べ ク ト ル の 計算 の 方法 に も 、 フ レ ー ム 予 測 と フ ィ ー ル ド 予 測が あ る 。 フ レ ー ム 予測で は 参照 画 像 を フ レ ー ム 構造 と し て扱 い 、 フ ィ 一 ル ド 予则で は 参 照画 像 を フ ィ 一 ル ド 構造 と し て (参照画 像が フ レ ー ム 構造 の 場 台 は 、 偶数 ラ イ ン あ る い は奇数 ラ イ ン の み の 画 像 と し て ) 扱 う 。
M P E G 1 規格で は 、 フ レ ー ム 構造 し か対応 し て い な い 。 そ の 場 台 の 動 き 補償予 測符号 化 の 伸長 回路 の 例 を F 1 G . 1 1 に 示 す。 F 1 G . 1 1 に 於 い て 、 加算器 2 が参照画 像 の 内 挿処理 を 行 う ブ ロ ッ ク で あ り 、 加算器 3 が復元処理 を 行 う ブ ロ ッ ク で あ る 。 参照画 像 メ モ リ は フ レ ー ム メ モ リ で あ る 。 F 1 G . 1 2 に 処理 の タ イ ミ ン グ チ ヤ 一 卜 を示す 随時 伝送 さ れ て く る 差分 デ ー タ に 対 し て 、 対応す る 参照 画 像 を随時 ア ク セ ス し 、 読 み だ し と 同時 に 内揷 を行 い 、 内 揷終了 と 同時 に 復元処理 を行 う 。 F 1 G . 1 2 に 示す よ う に 、 こ れ ら の 処理 は連続 し て 行 わ れ る の で 、 内 挿処理 と 復元処理 は 同時刻 に 行 わ れ る 。 差分 デ ー タ に つ い て は 、 演算器 6 で逆量子 化、 逆 コ サ イ ン 変換 な ど の 処理が行 わ れ、 そ こ で相 当数 の 遅れが生 じ る 。 そ の た め 、 そ れ ら の 処理 の 結果 の 出力 に 合 わ せ て 、 参照画 像 を ア ク セ ス す る タ イ ミ ン グ を 調整 し な け れば な ら な い 。
M P E G 2 規格 の 信号で伝送 さ れ る 画 像 に は 、 伝送 さ れ る 差分 の 画像構造、 予測方式 と も に フ レ ー ム と フ ィ ー ル ド の 両方 を と り 得 る 。 M P E G 2 の 様 な 画 像構造、 予 測方式 と も に フ レ ー ム と フ ィ ー ル ド の 両方 を と り 得 る 信 号 を扱 う 画 像圧縮伸長装置 に 於 い て は 、 復元処理 の 前 に フ レ ー ム / フ ィ ー ル ド変換装置 を 設 け て 、 差分 の 画像 の 構造 と 予測方式 を 合 わ せ な け れば な ら な い 。 そ の 場合 の 動 き 補償予測符号 化 の 伸長 回路 の 例 を F 1 G . 1 3 に 示す。 F I G . 1 3 に 於 い て 、 加算器 2 が参照画 像 の 内 挿処理 を行 う プ ロ ッ ク で あ り 、 加算器 3 が復元処理 を 行 う プ ロ ッ ク で あ る 。 ま た 、 中 間 メ モ リ で フ レ ー ム ノ フ ィ ー ル ド変換 を 行 う 。 参照画 像 メ モ リ は フ レ ー ム メ モ リ で あ る 。 フ レ — ム 全体 を 小画面 ( た と え ば 8 X 8 の ブ ロ ッ ク ) に 分割 し て処理す る 場合 の 処理 内 容を F I G . 1 4 の タ イ ミ ン グ チ ャ 一 ト を用 い て説明 す る 。 参照画 像 メ ΐ リ 5 力、 ら 、 小画 面 と そ の 周 辺画素 を 読み だ し 、 加算器 2 で 内挿 し 、 そ の 結果 を 中 間 メ モ リ 4 に 書 き 込む。 差分 デ ー タ は演算器 6 で逆量子 化、 逆 コ サ イ ン 変換 な ど の 処理が行 わ れ る 。 こ れ ら の 処理 は F 1 G . 1 4 の ( 1 ) の 期 間 に 同時 に 行 わ れ る 小画面 に つ い て そ れ ら の 処理が終了す る と 、 続 い て 、 中 間 メ モ リ 4 力、 ら 差分 デ ー タ と 同 じ フ レ ー ム フ ィ ー ル ド 構造で参照画 像 を読み だ し 、 外部か ら 入力 さ れ た 差分 デ 一 夕 と 加算器 3 で加算 さ れ、 復元処理が行 わ れ る 。 こ れ ら の 処理 は F I G . 1 4 の ( 2 ) の 期 間 に 同時 に 行わ れ る 。
こ の よ う に 中 間 メ モ リ を持 た ず、 差分 デ ー タ の 処理 に 合 わ せ て参照画像 を読み込ん で来て 内挿処理、 復元処理 を 行 う 画像伸 長装置で は 内 挿用 の 加算器 と 復元処理用 の 加算器 と が必.要で、 ハ ー ド ウ ェ ア 量が多量 な も の に な つ て い た 。 ま た 、 中 間 メ モ リ を持 っ て い て 、 フ レ ー ム ノ フ ィ ー ル ド変換処理 を 行 う 動 き 補償予測符号 化画像伸長装 置で は 、 ハ ー ド ウ エ ア の量が多量 な も の に な る ば力、 り で な く 、 中 間 メ モ リ の 前後の 加算器の 動作す る 時 間 は ォ ー バ ー ラ ッ プが無 く 、 一方が処理 を行 っ て い る 間 、 他方が 使用 さ れ な い 時 間が存在す る の で、 非常 に 効率 の 悪 い シ ス テ ム と な る と 言 う 欠点力《 あ っ た 。
—方 、 近年、 デ ィ ジ 夕 ル信号処理 に よ り デ ィ ジ 夕 ノレ符 号 の 圧縮 Ζ伸 長 を 行 な う 装置が開発 さ れて い る 。 特 に 符 号量が多 い デ ィ ジ タ ル画 像 を 伝送ノ蓄積す る た め に Μ Ρ E G , J P E G , H 2 6 1 な ど の 世界規格 が定 め ら れ、 こ れ ら の 規格 に 準 じ た 圧縮 伸 長装置が多 く 開発 さ れ て い る 。 こ れ ら の 規格 で は符号圧縮の 方法 と し て可変長符 号 と 時 間軸 方 向 の 相 関 を利用 し た 動 き 補償、 離散 コ サ イ ン 変換 を 利用 し て い る 。 復号装置 と し て は可変長符号 を 利用 す る こ と に 伴 い 、 画像 と 圧縮 さ れ た 符号 の 同期 を取 る た め に 圧縮 さ れて い る 入カ コ ー ド を一時貯 え て お く た め の コ ー ド入力 バ ッ フ ァ が必要 と な る 。 ま た 、 動 き 捕償 を利用 す る こ と に 伴 い 、 過去又 は未来 の フ レ ー ム 画 像 を 参照画 像 メ モ リ と し て保持 し て お く 必要が あ る 。 ま た 、 入力 符号 は 出 力 が ィ ン タ ー レ 一 ス ノ ン ィ ン 夕 一 レ ー ス に 関係 な く 2 次元 D C T を 行 な う プ ロ ッ ク 単位で送 ら れ て く る の で復号 し た 画像 を イ ン タ ー レ ー ス ま た は ノ ン ィ ン タ ー レ ー ス で 出 力 す る た め の フ レ ー ム メ モ リ が必要 と " る 。
す な わ ち 、 同期 を 取 る た め の コ ー ド入力 バ ッ フ ァ 、 動 き 補償用 参照画 像 フ レ ー ム メ モ リ 、 フ レ ー ム / フ ィ ー ル ド変換用 表示画像 フ レ ー ム メ モ リ の 3 種類 の メ モ リ 装置 が必要で あ る 。 デ コ ー ド装置 を 1 つ の L S I で構成す る と き 、 現在 の L S I の 集積度で は こ れ ら の メ モ リ は デ コ - ド装置 と 同 じ 1 つ の L S I に 集積す る こ と が難 し い た め 、 外部 メ モ リ と し て L S I の 外 に 設定 し な け れば な ら な い 。 し 力、 し 、 そ れ ぞれ の 用 途 に 别 々 の メ モ リ を割 り 当 て て い て は 配線数、 回路規模、 メ モ リ 量が大 き く な つ て し ま う の で外部 メ モ リ と し て 1 つ の メ モ リ を 上記参照画 像、 表示画像、 コ ー ド入力バ ッ フ ァ に分割 し て共用 し て 用 い る 方法が多 く 使われて い る 。
通常、 動画を圧縮 し た符号列デ コ ー ドす る 場合 に は入 力符号の コ ー ド入力バ ッ フ ァ へ の書 き 込み、 コ ー ド入力 バ ッ フ ァ 力、 ら の読み 出 し 、 動 き 補償 に と も な う 参照画像 の読み 出 し 、 デ コ ー ド後 に参照画面を再構成す る 参照画 像への書 き 込み、 デ コ ー ド し た画像を表示す る た め の表 示画像読み 出 し 、 の 5 種類の メ モ リ ア ク セ ス が発生す る が、 間断な く デー タ が流れて く る 入力符号の コ ー ド入力 バ ッ フ ァ へ の書 き 込み 、 常時画像を表示 し な ければな ら な い表示画像読み 出 し はバ ッ フ ァ があふれた り 、 な く な る こ と はで き ず、 ま た処理のオ ー バ ー へ ッ ドの関係か ら 参照画像の読み 出 し と 書 き 込み は一度行な つ た ら 途中で 止め る こ と も 難 し い。
1 つ の外部 メ モ リ に 対 す る こ れ ら の メ モ リ ア ク セ ス 要 求を処理す る た め、 従来 は 2 つ の方式が取 ら れて い る 。 1 っ は メ モ リ ア ク セ ス の必要が発生す る ご と に ア ク セ ス 要求をバ ス 調停装置 に送 り 、 バ ス調停装置 は各 ブ ロ ッ ク か ら 発生す る 要求信号を優先順位に し たが っ て調停を行 う 優先順位方法で あ る 。 ま た、 も う 1 つ の方式 と し て そ れぞれの要求を優先順位な し で順番 に処理す る ポ ー リ ン グ方式があ る 。 前者の方式の場合、 メ モ リ の バ ン ド幅 に 対 し メ モ リ ア ク セ ス が少な い と き に は問題 は な い メ モ リ ア ク セ ス 量が多 く な る と 優先順位が低い メ モ リ ァ ク セ ス は行われず待 ち状態が多 く な る 。 そ の た め処理の ォ 一 ノく 一 へ ッ ドカ《多 く な り 、 1 フ レ ー ム の 画像 の 復号力く 1 フ レ ー ム 表示時 間 に 行 な え な く な り 、 入力符号 と 出 力 画像 の 同期が取れ な く な っ て し ま う と い う 問題点が あ っ た 。 ま た 、 後者 の 方式 の 場合、 連続 し た メ モ リ ア ク セ ス が行 な え な い た め 、 外部 メ モ リ に D R A M を用 い た と き は ぺ — ジ 切 り 換 え の ォ 一 バ ー へ ッ ド力く多 く な り 、 メ モ リ バ ン ト 幅 の 不 足が発生す る 。
ま た 一方 、 動画 像 の 冗長性 と し て は 、
( 1 ) 空 間軸方 向 隣 の 画素 と よ く 似 て い る 、 人間 の 目 は钿 力、 い も の は よ く 見え な い 。
( 2 ) 時 間軸方 向 連続 し て い る 画像 は よ く 似 て い る
の 2 つ 力 あ る 。
こ の ( 1 ) , ( 2 ) の 冗長性 を圧縮す る た め 、
( 1 ) · あ る 大 き さ の ブ ロ ッ ク を単位 と し て 、 離散 コ サ イ ン 変換 ( D C T ) を お こ な い周 波数成分 に 変換 し 、 量子 化 を お こ な い高 い 周 波数成分 を削減す る 。
♦ 隣の ブ ロ ッ ク と の 差分 ( D P C M ) の み デ ー タ と す る
( 2 ) · あ る 大 き さ の ブ ロ ッ ク を単位 と し て 、 動 き 捕 償 を お こ な い 、 動 き べ ク ト ル と 誤差成分 の み を デ ー タ と す る 。
と い う 方法力く用 い ら れ る 。
こ れを 、 一般の 画 像 に 適用 す る と 、 比較的空間铀方向 の 冗長性 の 多 い 部分 (動 き の 激 し い 画像) は 動 き 補償が 適用 し に く い の で 、 空 間軸方法 の 冗長性 を圧縮す る た め D C T + 量子 化、 D P C M を お こ な う 。 こ の 場合 、 D P C M は 隣 の プ ロ ッ ク と 最 も 相関 の 高 い D C 成分 の み お こ な う 。
—方 、 比較的時間軸方向 の 冗長性 の 多 い 部分 ( あ ま り 動 い て い な い 画像 ) は動 き 補償を適用 し 、 動 き べ ク ト ノレ と 誤差成分 の み を デ ー タ と す る 。 ま た 、 さ ら に 空 間軸方 法 の 冗長性 を削減す る た め 、 誤差成分 に 対 し て も あ る 大 き さ の ブ ロ ッ ク 単位で D C T + 量子化を お こ な う 。 但 し 誤差成分 は 隣の プ ロ ッ ク と の 空間 的 な 相 関 は高 く な い と 考 え ら れ る た め D P C M は お こ な わ な い 。
以上 を ま と め る と 、
( 1 ) 動 き が激 し い 画像
D C T 係数 の D C 成分 D P C M
D C T 係数 の A C 成分 D C T + 量子 化 ( 2 ) 動 き の 少 な い 画 像
動 き 補償 + D C T + 量子化
と な る 。
以上の よ う に 符号 化 さ れ た デー タ を 画像へ復元す る 従 来の 画像処理装置 を 図 1 5 に 示す。 ま ず初 め に 、 符号 読回路 3 0 で可変長符号 の解読を行 い 、 フ ラ グ や D C T 係数 を求 め る 。 次 に 、 動 き 補償を行 っ た プ ロ ッ ク の 場合 は 、 逆量子 化 ( I Q ) · 逆離散 コ サ イ ン 変換 ( I D C T ) 回路 3 1 で差分 デ ー タ の D C T 係数 を I D C T し て差分 デ ー タ を求 め 、 同時 に 動 き べ ク ト ル再構成回路 3 2 で FI G . 1 6 に 示 さ れ る よ う な ア ル ゴ リ ズ ム で動 き べ ク ト ル を 再構成 し 、 そ の 動 き ベ ク ト ル を用 い て動 き 補償 回路 3 3 で画 像 メ モ リ 3 4 か ら 前方 ま た は後方 ま た は両方 の 画像 デ ー タ を取 り 出 し 、 加算回路 3 5 で差分 デ 一 夕 と 加算 し て復元画素 デ ー タ を 作成す る 。
動 き 補償 を 行 わ な い で画素 デ 一 夕 そ の も の を D C T し た ブ ロ ッ ク の 場合 は 、 D C T 係数 の D C 成分 の み D P C M処理 回路 3 6 で F1G. 1 7 に 示 さ れ る よ う な D P C M処 理 を行 い 、 他 の D C T 係数 と 併せ て I Q * I D C T 回路 3 1 で I D C T を行 い 復元画素 デ ー タ を作成す る 。 以上 が符号 化 さ れ た 画 像 デ ー タ を 画像 デ ー タ へ復元す る 従来 の 方法で あ る 。 な お 、 F 1 G . 1 6 及 び F 1 G . 1 7 で示 し た 各 コ マ ン ド の 意味 を FIG. 1 8 に 示す。
以上の よ う に 、 従来 の 中 間 メ モ リ を持 た ず、 差分 デ ー タ の 処理 に 合 わ せ て参照画像 を読み込ん で来 て 内 挿処理、 復元処理 を行 う 画像伸長装置で は 回路規模が多大 な も の に な っ て い た 、 さ ら に 、 中 間 メ モ リ を持 ち 、 フ レ ー ム ノ フ ィ ー ル ド変換を 伴 う 動 き 補償予測符号 化画 像伸長を行 う 画像伸長装置で は 、 加算器 の 回路規格が大 き く な り 、 し 力、 も 、 そ の 加算器が使用 さ れな い時間が存在す る と い う 問題点が あ っ た 。
ま た 、 前述 し た よ う に 、 従来の一般的 な バ ス 調停機能 で は 、 外部 メ モ リ へ ア ク セ ス す る 際 に 多大 な オ ー バ ー へ ッ ドが発生す る と い う 欠点力 あ っ た 。
さ ら に 、 上記 し た よ う に 、 動 き べ ク ト ル再構成 や D C T 係数 の D C 成分 の D P C M処理 は 、 解読 し た フ ラ グ を 用 い て F I G . 1 6 及 び F I G . 1 7 の よ う な 条件判断、 各種演 算 を行 う た め な ん ら か の シ ー ケ ン サ が必要で あ り 、 ハ ー ド ウ エ ア と し て大 き な も の と な る 。 ま た 、 い ろ い ろ な 規 格 や 応用 が考 え ら れ る 画像 デ ー タ の 圧縮 ♦ 伸 張技術で は フ ラ グ の 解釈 や演算の 仕方が異 な る 可能性が あ る た め上 記 シ ー ケ ン サ は ソ フ ト ゥ ヱ ァ 制御可能 な も の の 方が適応 性 · 拡張性が良 く な る が、 命令 メ モ リ な どが必要 と な り さ ら に ハ ー ド ウ ユ ア 量が大 き く な る 。 従来 の 画 像処理装 置 は 、 シ ー ケ ン サ 力く 2 つ も 必要 な た め 、 ハ ー ド ウ ェ ア 量 が多 く な る 傾 向 力《 あ っ た 。 発明 の 開示
本発明 は こ の 点 を鑑み て な さ れた も の で、 本発明 の 目 的 は 、 少 な い 回路規模で動 き 補償予測符号 化画像伸 長を 行 う 画像処理装置 を提供す る こ と に あ る 。 ま た 、 本発明 の 別の 目 的 と す る と こ ろ は、 外部 メ モ リ の 読み 出 し 書 き 込 み の 内部バ ッ フ ァ を用 途別 に も つ こ と に よ り 、 外部 メ モ リ への ア ク セ ス を効率良 く 行 な う こ と がで き る 画 像処理装置 を提供す る こ と に あ る 。 さ ら に 、 本発明 の 別 の 目 的 は、 ハ ー ド ウ ェ ア の 使用 効 率 の 良い 画像処理装置 を提供す る こ と で あ る 。
前記課題 を 解決す る た め に 本発明 に よ る 画 像処理装置 で は動 き 補償を行 う 処理領域分の画像 メ モ リ と 加算器か ら な り 、 時間的 に オ ー バ ー ラ ッ プの無い 内挿処理 と 復元 処理を同一の 加算器を時間多重で用 い て処理す る 。
ま た 、 実施例 と し て、 入力符号を解読 し 、 画像を復元 す る デ コ ー ド手段 と 、 外部か ら 入力 さ れ る 符号列を受け 取 っ て外部 メ モ リ に符号列を書 き 込む入力符号書 き 込み 手段 と 、 外部 メ モ リ に蓄積 さ れた符号列を読み 出 し 、 そ の符号列を前記デ コ ー ド手段 に転送す る 入力符号読み 出 し 手段 と 、 外部 メ モ リ に保存 さ れて い る 参照画像 を読み 出 し 、 前記デ コ ー ド手段で復元 さ れた 画像 と 加え る こ と に よ り 表示画 像を作成す る 画像再構成手段 と 、 前記画像 再構成手段で作成 さ れた画像を外部 メ モ リ に書 き 込む画 像書 き 込み手段 と 、 外部 に画像を表示す る た め に前記画 像書 き 込み手段に よ り 書 き 込 ま れた 画像を外部 メ モ リ よ り 読み 出す表示画像読み 出 し 手段 と 、 外部 メ モ リ と 前記 画像再構成手段、 前記画像書 き 込み手段、 前記入力符号 書 き 込み手段、 前記入力符号読み 出 し 手段、 前記表示画 像読み 出 し 手段を接続す る 接続手段 と 、 前記入力符号書 き 込み手段、 前記入力符号読み 出 し 手段、 前記表示画像 読み 出 し 手段が外部 メ モ リ か ら 読み 出 し た デー タ を そ れ ぞれ一時蓄積す る デー タ 蓄積手段 と 、 前記画 像再構成手 段、 前記画像書 き 込み手段、 前記入力符号書 き 込み手段 前記入力符号書 き 込み手段、 前記入力符号読み 出 し 手段 前記表示画 像読み 出 し 手段の 中か ら 前記接続手段 と 接続 す る 手段を選択す る メ モ リ バ ス調停手段 と で構成 し 、 前 記 メ モ リ バ ス 調停手段 は前述 し た 入力符号の コ ー ド入力 ノ ' ッ プ ア へ の 書 き 込 み 、 コ ー ド入力 バ ッ フ ァ 力、 ら の 読み 出 し 、 動 き 補償 に と も な う 参照画像 の 読み 出 し 、 デ コ 一 ド後 に 参照画像 を 再構成す る 参照画像へ の 書 き 込 み 、 デ コ ー ド し た 画 像 を 表示す る た め の 表示画 像読 み 出 し 、 の
5 種類 の メ モ リ ア ク セ ス を ア ク セ ス 優先順 位 の 高 い 入力 符号 の コ 一 ド 入力 バ ッ フ ァ へ書 き 込み 、 コ ー ド入力 バ ッ フ ァ 力、 ら の 読 み 出 し 、 デ コ ー ド し た 画 像 を 表示す る た め の 表示 画像読 み 出 し の 第 1 の グ ル ー プ と 優先順位 は 低い が連続 し て行 な い た い 動 き 捕償 に と も な う 参照画像 の 読 み 出 し 、 デ コ ー ド後 に 参照画像 を再構成す る 参照画像へ の 書 き 込 み の 第 2 の グ ル ー プ に 分 け 、 第 1 の グル ー プ 内 は ポ ー リ ン グ方式 に よ り あ る 決 ま っ た 期間だ け ア ク セ ス 権 を与 え 、 第 1 の グル ー プの 要求がすべて終了 し た と こ ろ で、 第 2 の グ ル ー プへ ア ク セ ス 権が移 り 、 第 2 の ダ ル — プへ ア ク セ ス 権が移 っ た と き 、 第 2 の グ ル ー プ 内 に ァ ク セ ス 要求力 あ る と そ の メ モ リ ア ク セ ス を行 な い 、 第 2 の グ ル ー プ 内 の メ モ リ ア ク セ ス は 1 度ア ク セ ス 権 を与 え る と そ の 一連の メ モ リ ア ク セ ス が終了す る ま で他の ァ ク セ ス 要求 は却下 さ れ る よ う に し た も の で あ る 。
さ ら に 、 他 の 実施例で は 、 離散 コ サ イ ン 変換 と 動 き 補 償 を用 い 、 動 き 補償 を行わ な い 部分の 離散 コ サ イ ン 変換 係数 の み 、 そ の 一部 を D P C M 化す る こ と に よ っ て 画像 デ ー タ を圧縮 ♦ 符号 化す る 方式の 符号 か ら 画 像 デ ー タ を 復元す る 装置 に お い て 、 離散 コ サ イ ン 変換 係数 の D P C M の 処理 と 動 き べ ク ト ル の 再構成 の 双方 の 処理 を行 う こ と がで き る プ ロ グ ラ ム 制御可能 な シ ー ケ ン サ を 備 え た も の で あ る 。
以上 の よ う な 構成 に よ り 、 本発明 で は 、 動 き 補償予測 符号 化画 像伸 長 を行 う 画 像処理装 置 の 加算器 の 回路規模 を ほ ぼ 1 2 に 削減す る こ と 力《で き る 。
ま た 、 上記実施例 に 於 て 、 コ ー ド入力 バ ッ フ ァ の 書 き 込 み用 バ ッ フ ァ 、 コ 一 ド入力 バ ッ フ ァ の 読 み 出 し 用 バ ッ フ ァ 、 表示 画像読 み 出 し 用 バ ッ フ ァ が そ れ ぞれ あ る 一定 量 に な っ た ら の そ れ ぞれ メ モ リ ア ク セ ス を メ モ リ バ ス 調 停装置 に 要求 し 、 参照画像 の 読み 出 し 、 参照画 像 の 書 き 込 み は 内 部 の 復号 に 伴 う 処理が参照画 像 の 読み 出 し 、 参 照画像 の 書 き 込 みが必要 に な っ た と こ ろ でバ ス 調停装置 に 要求す る 。 メ モ リ バ ス 調停装置 は コ ー ド入力 バ ッ フ ァ の 書 き 込み 、 コ ー ド入力 バ ッ フ ァ の 読み 出 し 、 表示画像 の 読み 出 し の メ モ リ ア ク セ ス 要求 は一定 の順序 で 要求 の 処理 を行 な う 。 参照画像 の 書 き 込み 、 参照画像 の 読 み 出 し は コ ー ド入力 バ ッ フ ァ の書 き 込み 、 コ 一 ド 入力 バ ッ フ ァ の 読み 出 し 、 表示画像読み 出 し の メ モ リ ア ク セ ス 要求 力《 な く な っ た と き 処理 を行 な う 。 こ の よ う な メ モ リ バ ス の 調停を 行 な う こ と に よ り 、 優先順位方式 と 比較 し た 場 合、 参照画像 の 読み 出 し 、 書 き 込み を連続 し て行 な う こ と がで き る た め参照画像 の 読み 出 し 、 書 き 込 み 後 に 行 な う べ き 処理 を速 や か に 行 な う こ と がで き る た め 、 処理 の オ ー バ ー ヘ ッ ド が少 な く な り 、 よ り 大 き な 画 像 ま で极 ぅ こ と がで き る 。 ま た 、 優先順位方式 と 比較 し た 場合 内部 バ ッ フ ァ を大 き く す る こ と な く D R A Mへ の ア ク セ ス に 伴 う ペ ー ジ 切 り 替 え を少な く す る こ と がで き 、 高価 な 高 速 の メ モ リ を 使用 す る こ と な く 、 よ り 大 き な 画 像 を 扱 う こ と 力 <で き る 。
さ ら に 、 上記他 の 実施例 に 基づ く 画 像処理装 置 は 、 解 読後 の 各種 フ ラ グ を 1 つ の ソ フ ト ウ ェ ア 制御可能 な シ ー ケ ン サ で処理 し て I Q * I D C T I ]路 ゃ動 き 補償回路 に
D C T 係数 や動 き べ ク ト ル を与 え る 構成で あ る 。 シ 一 ケ ン サ カ 1 つ で あ る た め 、 D C T 係数 の D C 成分 の D P C M処理 と 動 き べ ク ト ル の 再構成 は時分割で実行す る こ と に な る 。 し 力、 し 、 従来技術で説明 し た よ う に 、 D C T 係 数 の D C 成分が D P C Mで符号化 さ れて い る 場合 、 動 き 捕償 を行わ な か っ た ブ ロ ッ ク で あ る た め 、 動 き べ ク ト ル の 再構成 を 行 .う 必要が な く 、 ま た 動 き 補償 を行 っ た プ ロ ッ ク で は 、 D C T 係数 の D P C M化 さ れて い な い た め D P C M処理 の 必要が な い 。 従 っ て、 本発明 の 画像処理装 置 は D C T 係数 の D C 成分力 D P C M処理 と 動 き べ ク ト ルの 再構成 は時分割で お こ な う こ と に な る が、 処理時間 の 損失 は な く 、 シ ー ケ ン サ力 1 つ で あ る の で ハ ー ド ゥ エ ァ の 量を必要最小限 に す る こ と がで き る 。
図面 の 簡単 な 説 明
FIG. 1 は第 1 の 実施例 に か か る 画像処理装置 の 構成図 FIG. 2 は第 1 の 実施例 に 於 け る 第 1 の 変形例 に か 力、 る 画 像処理装置 の 処理 内容 を示す タ イ ミ ン グ チ ヤ 一 ト 。
F 1 G . 3 は第 1 の 実施例 に 於 け る 第 2 の 変形例 に か か る 画像処理装置 の 処理 内容 を 示す タ イ ミ ン グ チ ヤ 一 ト 。
FIG.4 ( a ) 乃至 FIG.4 ( d ) は第 1 の 実施例 に 係 わ る 内揷、 復元処理 の 過程 の 画像構造 を 説 明 す る 図。
FIG. 5 は 第 1 の 実施例 に 係 わ る 内揷 フ ィ ル 夕 の 例 を 示 す 図。
FIG. 6 は第 2 の 実施例 に 係 る 画像処理装置 の 構成 を示 す ブ ロ ッ ク 図 で あ る 。
F I G . 7 は第 3 の 実施例 に 係 わ る 画像処理装置 の 構成 図( F1G.8 は第 3 の 実施例 に 係わ る 画像処理装置 の プ ロ グ ラ ム 例。
FIG. 9 ( a ) 及 び F 1 G . 9 ( b ) は第 1 の 実施例 に 係 わ る 動 き べ ク ト ル の 精度を説明す る 図。
FIG. 1 0 ( a ) 及 び FIG. 1 0 ( b ) は第 1 の 実施例 に 係 わ る 画像 の フ レ ー ム 構造、 フ ィ ー ル ド構造を説 明 す る 図。
F I G . 1 1 は 従来 の フ レ ー ム / フ ィ 一 ノレ ド変換 の な い 画 像処理装置 の 構成 図。
F1G. 1 2 は 従来の フ レ ー ム フ ィ 一 ノレ ド変換 の な い 画 像処理装置 の 処理 内容 を示す タ イ ミ ン グ チ ヤ 一 ト 。
FIG. 1 3 は 従来 の フ レ ー ム Z フ ィ 一 ル ド変換 の あ る 画 像処理装置 の 構成 図。
FIG. 1 4 は従来 の フ レ ー ム / フ ィ ー ル ド変換 の あ る 画 像処理装置 の 処理 内容 を示す タ イ ミ ン グ チ ヤ 一 ト 。 FIG. 1 5 は従来の画像処理装置の構成図。
FIG. 1 6 は動 き べ ク ト ル再構成の ア ル ゴ リ ズ ム の説明 図。
F I G . 1 7 は D C T 係数の D C 成分の D P C M処理の ァ ル ゴ リ ズ ム の説明図。
FIG. 1 8 は卩 .8 及び卩10. 1 6 及び FIG. 1 7 で示 し た 各 コ マ ン ド の 意味を説明す る 図。 発明 を実施す る た め の最良の形態
以下、 図面を用 い て本発明 の実施例を説明す る 。
第 1 実施例
第 1 の実施例 に かか る 画像処理装置の 例を F I G . 1 に 示 す。 F1G. 1 に於い て、 1 は 内挿処理、 復元処理を行 う 加 算器、 4 は小画面 (た と え ば 8 X 8 の ブ ロ ッ ク ) を保持 し て お く 中間 メ モ リ 、 5 は 1 フ レ ー ム分の参照画像を保 持 し て お く フ レ ー ム メ モ リ であ る 。
第一の変形例の処理の タ イ ミ ン グチ ヤ 一 ト を F1G. 2 に 示す。 フ レ ー ム 全体を小画面 ( た と え ば 8 X 8 の プ ロ ッ ク ) に分割 し て処理す る 場合の処理内容を F1G. 2 の タ イ ミ ン グチ ャ ー ト を用 い て説明す る 。 5 の参照画像 メ モ リ か ら 、 小画面 と そ の周辺画素を読みだ し 、 1 の加算器で 内挿 し 、 そ の結果を 4 の 中間 メ モ リ に書 き 込む。 差分デ 一 夕 は 6 の演算器で逆 コ サ イ ン変換な どの処理が行われ る 。 こ れ ら の処理は F1G. 2 の ( 1 ) の期間 に 同時 に行わ れ る 。 小画面 につ い て そ れ ら の処理が終了す る と 、 続い て 、 4 の 中 間 メ モ リ か ら 参照画像 を読み だ し 、 外部 か ら 入力 さ れ た 差分 デ 一 夕 と 内挿処理で用 い た も の と 同 じ 1 の 加算器で加算 さ れ、 復元処理が行わ れ る 。 こ れ ら の 処 理 は 図 2 の ( 2 ) の 期 間 に 同時 に 行 わ れ る 。
第一 の 変形例 に よ る と 、 差分 デ ー タ の逆 コ サ イ ン 変換 な ど の 処理 の 間 に 参照画像 の 内 揷処理が行 わ れ、 同一 の 加算器 を 内 挿処理 と 復元処理 に 使 う の で 、 加算器 の 回路 規模が ほ ぼ 1 / 2 に 削減 さ れ、 無駄の 無 い 効率 的 な シ ス テ ム を 実現 で き る 。
第二 の 変形例 の 処理 の タ イ ミ ン グ チ ヤ 一 ト を F I G . 3 に 示す。 第二 の 変形例で は 4 の 中 間 メ モ リ を用 い て フ レ ー ム Z フ ィ ー ル ド変換 を行 う 。 第一の 実施例 と 同様 に 、 フ レ ー ム 全体 を 小画面 ( た と え ば 8 X 8 の ブ ロ ッ ク ) に 分 割 し て処理す る 場 台 の 処理 内容 を 図 3 の タ イ ミ ン グ チ ヤ ー ト を用 い て.説明す る 。 5 の 参照画 像 メ モ リ 力、 ら 、 小画 面 と そ の 周 辺画素 を読 み だ し 、 1 の 加算器で 内 揷 し 、 そ の 結果を 4 の 中 間 メ モ リ に書 き 込む。 差分 デ ー タ は 6 の 演算器で逆 コ サ イ ン 変換な ど の 処理が行わ れ る 。 こ れ ら の 処理 は F I G . 3 の ( 1 ) の 期間 に 同時 に 行 わ れ る 。 小画 面 に つ い て そ れ ら の 処理が終了す る と 、 続 い て 、 4 の 中 間 メ モ リ 力、 ら 差分 デ ー タ と 同 じ フ レ ー ム / フ ィ ー ル ド構 造で参照画像 を読 み だ し 、 外部か ら 入力 さ れ た 差分 デ ー 夕 と 内揷処理で用 い た も の と 同 じ 1 の加算器で加算 さ れ 復元処理が行 わ れ る 。 こ れ ら の処理 は F 1 G . 3 の ( 2 ) の 期 間 に 同時 に 行わ れ る 。 第二の変形例 に つ い て、 処理の詳細を、 FIG.4 及 び FI G . 5 を用 い て説明す る 。 こ の例の場台、 4 X 4 の ブ ロ ッ ク で処理を行 う 。 一 つ の動 き べ ク ト ノレ で 4 x 4 の ブ ロ ッ ク 全体の動 き 補償を行い、 4 X 4 の プ ロ ッ ク 内で フ レ 一 ム フ ィ ー ル ド変換行 う 。 動 き 補償 は縦横共 に 1 Z 2 精 度 と し 、 4 点の 内揷を行 う 。 FIG.4 の ( a ) は フ ィ 一 ル ド予測の時の参照画 像の構造を示 し てい る 。 a 1 か ら a 2 0 は画素 に 相 当す る 。 実線で結ばれた 画素 は第一 フ ィ 一 ル ド、 点線で結ばれた画素 は第二 フ ィ ー ル ド に相 当す る 。
F! G . 3 の ( 1 ) 期間で は、 参照画像 メ モ リ 力、 ら フ ィ 一 ル ド単位で a l , a 2 , a 3 , a 4 , a 5 , a 6 , a 7 a 8 , ♦ ♦ ♦ の順に読み 出 さ れて FIG. 5 に示す 内揷 フ ィ ル 夕 に入力 さ れ る 。 内挿フ ィ ル タ で は現画素 7 と 右 どな り の 画素 6 、 1 ラ イ ン 上の画素 1 と そ の右 ど な り の画素 2 で平均を取 る 。 内挿結果 は 中間 メ モ リ に書 き 込む。 中 間 メ モ リ の書 き 込み デー タ の構造を FIG.4 ( b ) に示す 以上の処理を FIG.4 ( a ) , ( b ) で示 さ れた記号を用 い て示す と 、
b l = ( a l + a 2 + a 6 + a 7 ) / 4 、 と な る 。 同様の処理を第一 フ ィ ー ル ド の b l 〜 ! 3 8 ま で 行い 、 続い て第二フ ィ ー ル ドの b 9 〜 b 1 6 ま で行 う 。
F 1 G . 3 の ( 2 ) の期間で は フ ィ ー ル ド単位で書 き 込 ま れた デー タ を フ レ ー ム単位で読み 出す。 すな わ ち 、 b a b 2 , b 3 , b 4 , b 9 , b l 〇 , b 1 1 , b 1 2 , b 5 , b 6 , ♦ ♦ ♦ の 順で読み だ し 、 フ レ ー ム 構造で伝送 さ れ た 差分 デ ー タ c l , c 2 , c 3 , c 4 , c 5 , c 6 , c 7 , c 8 , c 9 , c 1 0 , · · ♦ ( FIG.4 ( c ) ) カロ 算 さ れ、 復元 さ れ た d l , d 2 , d 3 , d 4 , d 5 , d 6 , d 7 , d 8 , d 9 , d 1 0 , · ♦ · ( FIG.4 ( d ) ) を得 る 。
d l = b l + c l , d 2 = ♦ - ♦
d 5 = b 9 + c 5 , d 6 = · · ·
中 間 メ モ リ 力、 ら デ ー タ を読み だ し て い る 間、 内揷 フ ィ ル 夕 で 加算が行 わ れ な い の で、 復元処理 の 加算 は F 1 G . 5 の 内揷 フ ィ ル 夕 の 加算器 を用 い る こ と がで き る 。
第二 の 変形例 に よ る と 、 差分 デ ー タ の 逆 コ サ イ ン 変換 な ど の 処理 の 間 に 参照画 像 の 内揷処理が行わ れ、 同一の 加算器 を 内 揷処理 と 復元処理 に 使 う の で 、 加算器 の 回路 規模 は削減す る こ と がで き る と と も に 、 中 間 メ モ リ で フ レ ー ム Z フ ィ ー ル ド の 変換 も 行わ れ る の で 、 メ モ リ の 增 加 も 無 く 、 予測方式が前方 と 後方 の 双方 向 の 予 測 の 場合 に は 回路規模 は従来例 の ほ ぼ 1 ノ 2 に 削減で き る 。
第 2 の 実施例
F I G . 6 は 、 第 2 の 実施例 の 画像処理装置 に 係 る 構成 を 示す ブ ロ ッ ク 図で あ る 。
F I G . 6 に お い て 、 こ の 復号装置 1 0 1 は デ コ ー ド処理 部 1 0 2 、 入力 符号書 き 込 み部 1 0 3 、 入力 符号読 み 出 し 部 1 0 4 、 表示 画 像読み 出 し 部 1 0 5 、 参照画 像再構 成部 1 0 6 、 参照画 像書 き 込 み 部 1 ◦ 7 、 メ モ リ バ ス 調 停部 1 0 8 、 バ ッ フ ァ 1 0 9 〜 1 1 1 、 ァ ド レ ス 発生部 1 1 2 、 メ モ リ バ ス 1 1 3 力、 ら 構成 さ れ る 。 ま た 、 こ の 復号装置 1 0 1 の 外部 に メ モ リ 装置 1 1 4 が接続 さ れて い る o
外部 か ら の 入力符号 は入力 符号書 き 込 み 部 1 0 3 に 入 力 さ れ、 バ ッ フ ァ 1 0 9 に 蓄積 さ れ る 。 バ ッ フ ァ 1 0 9 に 蓄積 さ れ た 符号が あ る 一定量以上 に な る と メ 乇 リ バ ス 調停部 1 0 8 に 書 き 込 み 要求 を送 り 、 メ モ リ バ ス 調停部 1 0 8 力、 ら メ モ リ ア ク セ ス 許可信号が返 っ て き た ら メ モ リ バ ス 1 1 3 と ァ ド レ ス 発生部 1 1 2 に よ り メ モ リ 装置 1 1 4 へバ ッ フ ァ 1 0 9 内 の デ ー タ を決 ま っ た 量だ け 書 き 込 む。
符号読 み 出 し 部 はバ ッ フ ァ 1 1 0 の 蓄積量が あ る 一定 量以下 に な る と メ モ リ バ ス 調停部 1 0 8 に 読み 出 し 要求 を 送 り 、 メ モ リ バ ス 調停部 1 0 8 力、 ら メ モ リ ア ク セ ス 許 可信号が返 っ て き た ら メ モ リ バ ス 1 1 3 と ァ ド レ ス 発生 部 1 1 2 に よ り メ モ リ 装置 1 1 4 力、 ら デ一 夕 を 決 ま っ た 量 だ け読み込 み 、 バ ッ フ ァ 1 1 0 へ蓄積す る 。 符号読 み 出 し 部 は ノく ッ フ ァ 1 1 ◦ に 蓄積 さ れて い る デ ー タ を デ コ 一 ド処理部 1 ◦ 2 に 順次入力す る 。
参照画像再構成部 1 0 6 は デ コ ー ド処理部 1 ◦ 2 で画 像が復号 さ れ た ら 、 メ モ リ バ ス 調停部 1 0 8 に 読 み 出 し 要求 を送 り 、 メ モ リ バ ス 調停部 1 0 8 力、 ら メ モ リ ァ ク セ ス 許可信号が返 っ て き た ら メ モ リ バ ス 1 1 3 と ァ ド レ ス 発生部 1 1 2 に よ り メ モ リ 装置 1 1 4 か ら 必要な デ ー タ を すベ て読 み 込 み 、 読 み込ん だ デ ー タ と デ コ ー ド処理部 で復号 さ れ た デ ー タ と で予測 に 伴 う 画 像 の 再構成 を行 な フ 0
参照画像書 き 込 み 部 1 ◦ 7 は参照画 像再構成部で 画像 デ ー タ が再構成 さ れ た ら 、 メ モ リ バ ス 調停部 1 0 8 に書 き 込 み 要求 を送 り 、 メ モ リ バ ス 調停部 1 0 8 力、 ら メ モ リ ア ク セ ス 許可 信号が返 っ て き た ら メ モ リ バ ス 1 1 3 と ァ ド レ ス 発生部 1 1 2 に よ り メ モ リ 装置 1 1 4 に 再構成 さ れ た 画像 デ ー タ を書 き 込 む
表示 画像読 み 出 し 部 1 0 5 は バ ッ フ ァ 1 1 1 に 蓄積 さ れ た デ ー タ を読み 出 し 、 必要 な 処理を施 し て外部 に 出 力 す る 。 ノく ッ フ ァ 1 1 1 に 蓄積 さ れ た デ 一 夕 が め る ~ ' At 以下 に な る と メ モ リ バ ス 調停部 1 〇 8 に s¾ 出 し 要求を 送 り 、 メ モ リ バ ス 調停部 1 0 8 力、 ら メ モ リ ア ク セ ス 許可 信号が返 っ て き た ら メ モ リ と ア ド レ ス 発生部 1 1 2 に よ り メ モ リ 装置 1 1 4 力、 ら デ ー タ を 決 ま っ た 量 だ け 読み込 み 、 バ ッ フ ァ 1 1 1 へ蓄積す る 。
メ モ リ バ ス 調停部 は表示画像読み 出 し 部 1 0 5 、 入力 符号書 き 込 み部 1 0 3 、 入力符号読 み 出 し 部 1 〇 4 の メ モ リ ア ク セ ス 要求 は一定 の 順序で要求 の 処理 ¾ 仃 な う 。 参照画像 の 書 き 込 み参照画像 の 読み 出 し は 表示 画像読み 出 し 部 1 0 5 、 入力 符号書 き 込み部 1 〇 3 、 入力 符号読 み 出 し 部 1 0 4 の メ モ リ ア ク セ ス 要求が な く な っ た と き に メ モ リ ア ク セ ス 許可信号 を発生す る
こ の よ う な 構成 に よ り 、 表示画像読み 出 し 部 1 0 5 、 入力符号書 き 込 み部 1 0 3 、 入力符号読み 出 し 部 1 0 4 の バ ッ フ ァ 1 〇 9 〜 1 1 1 の デ ー タ は空 ま た は い っ ぱ い に な っ て デ ー タ が途切れて処理が中断す る よ う な こ と は な く 、 加 え て参照画像 の書 き 込み 、 参照画像 の 読 み 出 し を 連続 し て 行 な う こ と がで き る た め 、 デ コ ー ド処理 と メ モ リ ア ク セ ス に 伴 う オ ー バ 一 ヘ ッ ド を少 な く す る こ と が で き る 。
第 3 の実施例
F 1 G . 7 に 、 第 3 の 実施例 の 画像処理装置 の 実施例 を示 す。 こ の 画 像処理装置 は 、 1 つ の ソ フ ト ゥ ユ ア 制御可能 な シ ー ケ ン サ で解読後 の 各種 フ ラ グを処理 し て I Q * I D C T 回路 や動 き 補償用 回路 に D C T 係数 や動 き べ ク ト ル を与 え る 構成で あ る 。
ま ず初 め に 、 符号解読回路 1 0 で可変長符号 の解読を お こ な い 、 フ ラ グ や D C T 係数 を求 め る 。 次 に 、 動 き 補 償 を お こ な っ た ブ ロ ッ ク の 場合 は 、 I Q · I D C T 回路 1 1 で差分 デ ー タ の D C T 係数 を I D C T し て差分 デ 一 夕 を求 め 、 同時 に プ ロ グ ラ ム で制御可能 な シ ー ケ ン サ 1 2 で動 き べ ク ト ル を再構成 し 、 そ の 動 き べ ク ト ル を用 い て動 き 補償 回路 1 3 で画像 メ モ リ 1 4 か ら 前方 ま た は後 方 ま た は両方 の 画像 デ ー タ を取 り 出 し 、 加算 回路 1 5 で 差分 デ ー タ と 加算 し て復元画素 デ ー タ を 作成す る 。 動 き 補償を お こ な わ な い で画素 デ ー タ そ の も の を D C T し た プ ロ ッ ク の 場合 は 、 D C T 係数 の D C 成分 の み シ 一 ゲ ン サ 1 2 で D P C M処理 を お こ な い 、 他の D C T 係数 と 併 せ て I Q · I D C T 回路 1 1 で I D C T を お こ な い 復元 画素 デ ー タ を 作成す る 。 第 3 の 実施例 の 画 像処理装置 に 用 い て い る シ ー ケ ン サ 1 2 は 、 F 1 G . 8 に 示す よ う な プ ロ グ ラ ム が実行で き 、 動 き べ ク ト ル の 再構成 や D C T 係数 の D C 成分 の D P C M処理を時分割で実行で き る 。
以上が符号 化 さ れ た 画 像 デ ー タ を 画 像 デ ー タ へ復元す る 第 3 の 発明 の 画 像処理装置 の 動作で あ る 。 産業上 の 利用 可能性
以上、 詳細 に 記載 し た よ う に 本発明 に よ る と 、 動 き 補 償予測符号 化 画 像伸 長 を行 う 画 像処理装置 の 加算器 の 回 路規模 を ほ ぼ 1 Z 2 に 削減す る こ と がで き 、 無駄 の 無 い 効率 的 な シ ス テ ム を 構成で き る 。
ま た 、 上記実施例 に 於て は 、 外部 メ モ リ の 読 み 出 し ノ 書 き 込み の 内部バ ッ フ ァ を用 途別 に 持 ち 、 参照画 像 の読 み 出 し Z書 き 込 み を連続 し て行 な っ て い る 。 こ れ に よ り 外部 メ モ リ へ の ア ク セ ス に 伴 う オ ー バ ー へ ッ ド が増大す る こ と な く 、 メ モ リ ア ク セ ス を効率良 く 行 な う こ と がで き る 。
さ ら に 、 別 の 上記実施例 に 於て は 、 シ ー ケ ン サ力 1 つ の 構成で あ る の で、 ド ウ ヱ ァ を 必要最小 に 抑 え る こ と がで き る 。 さ ら に 様 々 な 規格 や応用 に 対 し て 、 ハ ー ド ウ ェ ア の 増加 な し に 、 ソ フ ト ウ ヱ ァ の 開発の み で対応す る こ と がで き 、 ド ウ ヱ ァ の 使用 効率 の よ い 画像処理 装置 を提供す る こ と がで き る 。

Claims

請求の範囲
1 . 動 き 補償予測符号化画像伸長を行 う 画像処理装置で あ っ て、 動 き 捕償を行 う 処理領域分の 画像 メ モ リ と 加算 器か ら な り 、 前記加算器で、 1 Z 2 精度の動 き べ ク ト ノレ か ら 参照画像を構成す る 際の 内挿計算を行い 、 そ の結果 を前記画像 メ モ リ に記憶 し 、 同一の前記加算器で参照画 像 と 差分デ一 タ と の加算処理 も 時分割で処理す る こ と を 特徴 と す る 画像処理装置。
2 . フ レ ー ム Z フ ィ ー ル ド変換を伴 う 動 き 捕償予測符号 化画像伸長を行 う 画 像処理装置であ っ て、 動 き 補償を行 う 処理領域分の画像 メ モ リ と 加算器か ら な り 、 前記加算 器で、 1 ノ 2 精度の動 き ベ ク ト ルか ら 参照画像を構成す る 際の 内挿計算を行い 、 前記画像 メ モ リ で フ レ ー ム フ ィ — ル ド変換を行い 、 同一の前記加算器で参照画像 と 差 分デー タ と の加算処理 も時分割で処理す る こ と を特徴 と す る 画像処理装置。
3 . 入力符号を解読 し 、 画像を復元す る デ コ ー ド手段 と 外部か ら 入力 さ れ る 符号列を受け取 っ て外部 メ モ リ に 符号列 を書 き 込む入力符号書 き 込み手段 と 、
外部 メ モ リ に蓄積 さ れた符号列を読み 出 し 、 そ の符号 列を前記デ コ ー ド手段に転送す る 入力符号読み 出 し 手段 と 、
外部 メ モ リ に保存 さ れて い る 参照画像を読み 出 し 、 前 言己デ コ ー ド手段で復元 さ れた画像 と 加え る こ と に よ り 表 示画 像を作成す る 画 像再構成手段 と 、
前記画 像再構成手段で作成 さ れた画像を外部 メ モ リ に 書 き 込む画像書 き 込み手段 と 、
外部 に 画 像 を表示す る た め に前記画像書 き 込み手段 に よ り 書 き 込 ま れた 画 像を外部 メ モ リ よ り 読み 出す表示画 像読み 出 し 手段 と 、
外部 メ モ リ と 前記画像再構成手段、 前記画像書 き 込み 手段、 前記入力符号書 き 込み手段、 前記入力 符号読み 出 し 手段、 前記表示画 像読み 出 し 手段を接続す る 接続手段 と 、
前記入力符号書 き 込み手段、 前記入力符号読み 出 し 手 段前記表示画 像読み 出 し 手段が外部 メ モ リ か ら 読み 出 し た デー タ を そ れぞれ一時蓄積す る デー タ 蓄積手段 と 、 前記画像再構成手段、 前記画像書 き 込み手段、 前記入 カ府号書 き 込み手段、 前記入力符号前記入力符号書 き み 手段、 前記入力符号読み 出 し 手段、 前記表示画像読み 出 し 手段の 中か ら 前記接続手段 と 接続す る 手段を選択す る メ モ リ バ ス 調停手段 と を備え、
前記 メ モ リ バ ス 調停手段は前述 し た入力符号書 き 込み 入力符号読み 出 し 、 画像再構成に伴 う 読み 出 し 、 画 像書 き 込み 、 表示画像読み 出 し 、 の 5 種類の メ モ リ ア ク セ ス を ア ク セ ス 優先順位の高 い入力符号書 き 込み、 入力 符号 読み 出 し 、 表示画像読み 出 し の第 1 の グルー プ と 優先順 位 は低いが連続 し て行な い た い画像再構成 に 伴 う 読み 出 し 、 画像書 き 込み、 の第 2 の グル ー プに分 け、 第 1 の グ ル ー プ 内 は ポ ー リ ン グ方式 に よ り あ る 決 ま つ た 期間 だ け ア ク セ ス 権 を与 え 、 第 1 の グ ル ー プ の 要求がすべて終了 し た と こ ろ で 、 第 2 の グ ル ー プへ ア ク セ ス 権が移 り 、 第 2 の グ ル ー プへ ア ク セ ス 権が移 っ た と き 、 第 2 の グ ノレ 一 プ 内 に ア ク セ ス 要求力く あ る と そ の メ モ リ ア ク セ ス を行 い 、 第 2 の グ ル ー プ内 の メ モ リ ア ク セ ス は 1 度 ア ク セ ス 権を 与 え る と そ の 一連 の メ モ リ ア ク セ ス が終了す る ま で他 の ア ク セ ス 要求 は却下 さ れ る よ う に し た こ と を特徴 と す る 画 像処理装置。
4 . 離散 コ サ イ ン 変換 と 動 き 補償 を用 い 、 動 き 補償 を お こ な わ な い 部分の離散 コ サ イ ン 変換係数 の み 、 そ の 一部 を D P C M 化す る こ と に よ っ て画 像 デ ー タ を圧縮 ♦ 符号 化す る 方式の 符号か ら 画像 デ ー タ を 復元す る 装置 に お い て 、 離散 コ サ イ ン 変換係数 の D P C M の 処理 と 動 き べ ク ト ル の 再構成 を同一の 処理回路で お こ な う こ と を特徴 と す る 画像処理装置。
PCT/JP1994/001029 1993-06-28 1994-06-27 Processeur d'images Ceased WO1995001053A1 (fr)

Priority Applications (3)

Application Number Priority Date Filing Date Title
DE69425720T DE69425720T2 (de) 1993-06-28 1994-06-27 Bildverarbeitungsgerät
KR1019950700678A KR100252838B1 (ko) 1993-06-28 1994-06-27 화상처리 장치
EP94918576A EP0660608B1 (en) 1993-06-28 1994-06-27 Image processor

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP5157149A JPH0723389A (ja) 1993-06-28 1993-06-28 画像処理装置
JP5/157149 1993-06-28

Publications (1)

Publication Number Publication Date
WO1995001053A1 true WO1995001053A1 (fr) 1995-01-05

Family

ID=15643252

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP1994/001029 Ceased WO1995001053A1 (fr) 1993-06-28 1994-06-27 Processeur d'images

Country Status (6)

Country Link
US (1) US6035322A (ja)
EP (2) EP0660608B1 (ja)
JP (1) JPH0723389A (ja)
KR (1) KR100252838B1 (ja)
DE (2) DE69431572T2 (ja)
WO (1) WO1995001053A1 (ja)

Families Citing this family (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR100230251B1 (ko) * 1995-12-13 1999-11-15 윤종용 동영상 부호화장치에 있어서 신호처리방법 및 회로
JP3535297B2 (ja) 1996-01-26 2004-06-07 ローム株式会社 画像データ復号方法およびこの方法を用いた画像データ復号装置
JP4146027B2 (ja) * 1999-04-05 2008-09-03 株式会社岩根研究所 情報変換システム
BRPI0303336B8 (pt) 2002-07-02 2020-06-02 Matsushita Electric Industrial Co Ltd método de derivação do vetor de movimento, método de codificação de imagem animada e método de decodificação de imagem animada
TW591407B (en) * 2003-03-21 2004-06-11 Via Tech Inc Method for a display controller to access data stored in a system memory of a computer device
US8499337B1 (en) 2004-10-06 2013-07-30 Mcafee, Inc. Systems and methods for delegation and notification of administration of internet access
JP2007293606A (ja) * 2006-04-25 2007-11-08 Matsushita Electric Ind Co Ltd 画像処理装置
KR101228110B1 (ko) * 2006-07-26 2013-01-31 삼성전자주식회사 영상 보간 방법 및 장치
CN103108175A (zh) * 2011-11-09 2013-05-15 深圳市德赛微电子技术有限公司 多模视频解码器中运动补偿结构

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6239920A (ja) * 1985-08-15 1987-02-20 Nec Corp 画像信号の動き補償フレ−ム間予測符号化・復号化方法とその装置

Family Cites Families (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4710813A (en) * 1981-06-04 1987-12-01 Compression Labs, Inc. Low bandwidth video teleconferencing system and method
US5379351A (en) * 1992-02-19 1995-01-03 Integrated Information Technology, Inc. Video compression/decompression processing and processors
JP2521145B2 (ja) * 1989-02-20 1996-07-31 日本ビクター株式会社 動き補償予測符号化復号化方式
US5036493A (en) * 1990-03-15 1991-07-30 Digital Equipment Corporation System and method for reducing power usage by multiple memory modules
GB2252002B (en) * 1991-01-11 1995-01-04 Sony Broadcast & Communication Compression of video signals
CA2062200A1 (en) * 1991-03-15 1992-09-16 Stephen C. Purcell Decompression processor for video applications
US5212742A (en) * 1991-05-24 1993-05-18 Apple Computer, Inc. Method and apparatus for encoding/decoding image data
JPH04354287A (ja) * 1991-05-30 1992-12-08 Sony Corp 画像補間回路
KR0160610B1 (ko) * 1992-04-07 1999-01-15 강진구 가변장부호화에 의한 영상압축방법과 신장방법 및 그 장치
CA2096584A1 (en) * 1992-05-28 1993-11-29 Frank H. Liao Variable length code decoder for video decompression operations
US5289548A (en) * 1992-06-30 1994-02-22 Loral Aerospace Corp. Compression and reconstruction of radiological images
JP2725544B2 (ja) * 1992-11-12 1998-03-11 日本電気株式会社 Dct及び逆dct演算装置並びにその演算方法
US5367629A (en) * 1992-12-18 1994-11-22 Sharevision Technology, Inc. Digital video compression system utilizing vector adaptive transform
JP3190755B2 (ja) * 1993-01-22 2001-07-23 株式会社東芝 復号装置

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS6239920A (ja) * 1985-08-15 1987-02-20 Nec Corp 画像信号の動き補償フレ−ム間予測符号化・復号化方法とその装置

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
HIROSHI YASUDA (auther), "International Standard of Multi-Media Coding", June 30, 1991 (30.06.91), Maruzen, pages 18 to 23, 108 to 114. *
See also references of EP0660608A4 *

Also Published As

Publication number Publication date
DE69431572T2 (de) 2003-06-26
KR950703258A (ko) 1995-08-23
EP0896478B1 (en) 2002-10-16
DE69425720T2 (de) 2001-04-12
EP0660608B1 (en) 2000-08-30
DE69425720D1 (de) 2000-10-05
EP0660608A1 (en) 1995-06-28
EP0660608A4 (en) 1997-04-02
DE69431572D1 (de) 2002-11-21
EP0896478A2 (en) 1999-02-10
JPH0723389A (ja) 1995-01-24
US6035322A (en) 2000-03-07
KR100252838B1 (ko) 2000-04-15
EP0896478A3 (en) 1999-09-15

Similar Documents

Publication Publication Date Title
US5576765A (en) Video decoder
CN1196340C (zh) 用于数字视频运动补偿的搜索方法和处理器
US6571017B1 (en) Method and apparatus for digital image coding and decoding, and data recording media
JPH06326996A (ja) 圧縮されたビデオデータをデコードする方法及び装置
JPH08275160A (ja) 離散余弦変換方法
JP2001292451A (ja) 動画像信号圧縮装置及び方法
WO2004059980A1 (en) Method and apparatus for encoding and decoding stereoscopic video
JPH06225292A (ja) イメージデコーデングシステムのためのモジュールメモリ
JPH08275149A (ja) データ符号化方法
WO1995001053A1 (fr) Processeur d&#39;images
JP2006279573A (ja) 符号化装置と方法、ならびに復号装置と方法
JPH07240844A (ja) 画像データ処理装置および画像データ処理方法
JP2006262188A (ja) 画像符号化装置、及び画像復号化装置
JP2007523512A (ja) 量子化雑音伝搬現象を抑制するための最適な空間−時間変換
JPH07177514A (ja) 動画像圧縮装置
US5805483A (en) Method of converting data outputting sequence in inverse DCT and circuit thereof
JPH0730903A (ja) 画像処理用メモリ集積回路
JP3798432B2 (ja) ディジタル画像を符号化および復号化する方法および装置
JP2001045493A (ja) 動画像符号化装置、動画像出力装置、及び記憶媒体
JP3401761B2 (ja) 動画像圧縮符号化・復号化方法、動画像圧縮符号化・復号化装置、動画像符号化伝送方法、動画像符号化伝送システムおよび動画像圧縮符号化・復号化プログラムを記録した記録媒体
WO2006038679A1 (ja) 動画像符号化の装置、方法及びプログラムと、動画像復号の装置方法及びプログラム
JPH0622291A (ja) 解像度変換可能な復号方法と装置
JPH11155147A (ja) 画像再生方法、画像符号化装置及び画像符号化方法
JPH10150665A (ja) 予測画像の作成方法及び画像符号化方法及び画像符号化装置
JP2000115777A (ja) 画像処理方法および画像処理装置

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref country code: US

Ref document number: 1994 379562

Date of ref document: 19940627

Kind code of ref document: A

Format of ref document f/p: F

AK Designated states

Kind code of ref document: A1

Designated state(s): KR US

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH DE DK ES FR GB GR IE IT LU MC NL PT SE

WWE Wipo information: entry into national phase

Ref document number: 1994918576

Country of ref document: EP

121 Ep: the epo has been informed by wipo that ep was designated in this application
WWP Wipo information: published in national office

Ref document number: 1994918576

Country of ref document: EP

WWG Wipo information: grant in national office

Ref document number: 1994918576

Country of ref document: EP