JPH0244968A - Still picture transmitter - Google Patents

Abstract

PURPOSE:To shorten the time till the display of the first still picture by sending stepwise the still picture information so as to be made accurate gradually. CONSTITUTION:The transmitter is provided with an input frame memory storing an input vector signal of a pre-processing circuit 1 at a sender section, a difference vector signal generating circuit 10, an index data control circuit 11, a decoding circuit 12 and a reproducing signal generating circuit 13. Then the inter-frame vector quantizing coding system is adopted for the transmission of a still picture and the transmission information is compressed at the 1st vector signal readout to send the still picture is sent stepwise. Thus, the still picture is displayed roughly to the reception side in a short time from the transmission start command input and the content of picture is sent quickly to the user.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a still image transmission device, particularly to an improvement of a still image transmission device applying a vector quantization method in image communication such as a video conference or a video telephone. .

[Prior Art] In general, image transmission used for image communication such as video conferencing or video telephone calls involves a huge amount of image information, but due to line speed and transmission cost, etc. Compression encoding devices that reduce (compress) the amount of image information have been put into practical use.

Among these methods, the vector quantization method is known as a method for compressing the amount of information with a high degree of compression.

Hereinafter, a conventional still image transmission device will be explained based on the drawings.

FIG. 7 shows a basic block configuration of a still image transmission apparatus to which the tree search vector quantization method is applied, and FIG. 8 shows a detailed block configuration of a conventional tree search vector quantization section.

In FIGS. 7 and 8, a blocking unit (1) converts an image input signal (100) from A/D, blocks adjacent pixels on the image one by one, and generates a dimensional vector signal for each block. This is a circuit that creates (101).

On the other hand, when vector signal patterns are arranged in this structure as shown in FIG. It consists of (21), (22), (23), and (24).

Then, the main search circuit (21) in the - stage is a codebook (210), (21) that stores the vectors in the first stage of the segmentation stage.
1), the vector signal (101) and the codebook (2
10) Distortion calculation units (212) and (213) that perform distortion calculations with the pattern vectors output one by one from (211) as shown in the following equation, and a comparison that selects the pattern vector with the smaller distortion value. a codebook address register (215) that generates an address signal (2100) for the next stage (second stage) codebook based on the output signal of the comparator (214); ) to the next stage (second stage) tree search circuit (220)
and a latch circuit (216) that latches the output to the codebook address signal (2200) while generating the codebook address signal (2200).

Vectors are stored in the codebooks (210) and (211), respectively, as shown in FIG. 10(a). Tree search circuit at each stage (22) (23), (24)
has the same configuration as the -th stage tree search circuit (21).

Therefore, the vector that most closely approximates the vector signal (101) can be selected from the final stage pattern vector group (yoooo-yllll) shown in FIG. 9 by four stages of comparison operations. The final codebook address is output as index data (102).

For example, considering only brightness information, when encoding normal image information on a pixel basis, 8 bits are required per pixel. However, when one block is composed of 4×4 pixels, when vector quantization is performed as described above, the 4-bit information of the address of the pattern number becomes the information of one block. Therefore, 1 block (4X4
-It is possible to reduce information by 8×16-4=124 bits per 166 pixels, which is a very significant effect.

The index data (102) is sent to the transmission buffer (3), stored frame by frame in the transmission buffer (3), and sent out to the communication line frame by frame.

Next, the signal flow will be explained.

First, the image input signal (100) is input to the blocking unit (1
), the signal is converted into blocks, and converted into a vector signal (101).

Then, the vector signal (101) is quantized in the main search vector quantization section (2) as follows.

First, the vector signal (
101) and the vectors yo and y stored in the codebooks (210) and (211). ,dl
is determined by the distortion calculation unit (212) as shown in the following equation.

d-Σ IXk-Ykl (For absolute value child) However,
Xk: the first (component) of the input vector signal (103) to the codebook address register (215).For example, d.

is small, the codebook address register (21
5) sends “0”.

Then, the codebook address register (216) sets “o o o o” as the codebook address for the next stage.
” is output.

Next, in the two-stage book search circuit (22), the codebook (2
20) and (221) respectively address “o o o
o” vector, yoo and yol are the distortion calculation unit (2
22) and (223).

Then, the distortion calculation units (222) and (223) calculate the distortion value with respect to the vector signal (101), and a vector register for the third stage is generated.

Similarly, the three-stage main search circuit (23) and the four-stage main search circuit (24) perform distortion calculation comparisons, and select the vector that most closely approximates the vector signal (101) from the vector group of the fourth segmentation stage. The address of is selected and the index data (
102), it is sent to the transmission buffer (5) and temporarily stored for each frame.

Next, reception will be explained.

The decoding unit (6) receives the index data (102) from the communication line, and decodes the codebook (6) using the received index data (102) as address information.
From a), the pattern vector signal is converted into a decoded signal (106
).

At this time, the codebook (6a) of the receiving section includes a group of pattern vectors in the final stage (fourth stage in this example).

The decoded signal (106) is then D/A converted to generate a reproduced signal (107), which is displayed on the image display section.

Therefore, the more types of pattern vector groups there are, the more it becomes possible to transmit an image that approximates the vector signal, and the longer the data length of the index data, the more precise the image transmission becomes.

[Problem to be solved by the invention] In general, still images require finer image transmission.

Therefore, in the case of still image transmission, the amount of transmitted information is much larger than in the case of moving image transmission.

In addition, in the conventional still image transmission device, the transmitting unit stores index data for one frame in a transmission buffer, sends the index data to the communication line for each frame, and the receiving unit stores the index data in the final stage in the codebook. The index data received via the communication line was decoded and displayed after all data for one frame was received.

Furthermore, since the communication line speed is low compared to the large amount of information to be transmitted for still images, there is a problem in that it takes time from when an instruction to start image transmission is input until the image is displayed on the receiving side device.

Furthermore, even if the inter-frame difference n-coding method, which has a high degree of information compression and is applied to video transmission, is applied as is, the information in the first frame is transmitted by intra-frame encoding, so the transmission information in the first frame is The amount is not much different from that of conventional still image encoding methods, but there is a problem in that it takes time from inputting an image transmission instruction to displaying it on the receiving device.

The still image transmission device according to the present invention has been made to solve the above problems, and aims to shorten the time from the start of image data transmission to display without deterioration of image quality.

[Means for Solving the Problems] In order to achieve the above object, a still image transmission device according to the present invention applies an inter-frame vector quantization encoding method with a high degree of compression to still image transmission, and furthermore, The amount of transmitted encoded information in a frame (intraframe encoding) is reduced.

That is, the device reads a still image signal, A/D converts the image signal to generate a pixel signal, and collects the pixel signals into a plurality of blocks according to a predetermined rule to generate n blocks.
A preprocessing circuit that generates a vector signal Vj (where J indicates a block number) corresponding to each block j and stores each vector signal;
This circuit repeatedly reads vector signals of 1 frame worth) at predetermined intervals and outputs difference vectors corresponding to each of the vector signals V, and each difference vector signal D V 1, j (J- 1 to n), each of the vector signals V. (j=1 to n), and outputs the differential vector signal DV. during subsequent reading. ．． (i=1, J 2, 3, . . . , j=1 to n) as the reproduced signal group V. , (j=1 to n) 1-1
, J, and a codebook having pattern vectors arranged in this structure.
The difference vector signal D V outputted from the difference vector signal generation circuit. j manually, and the difference vector signal DV. .

A pattern vector signal that is most similar to IJ is selected from the codebook, and index data 1. ,
According to the actual search quantization coding 1 and 3 circuits that output
1. , the number of bits 1. becomes larger 1, J
1, each of the index data ■,.

1, from the most significant bit of J. Transmit bit data index data T1. , and the input/index data control circuit which outputs the input/index data T1. , a transmission control circuit that transmits or receives 1 frame worth of 1. , 1, J A decoding circuit that reads a corresponding pattern vector signal from the codebook and generates and outputs a decoded vector signal pv,j of the J-th block at the first reading from the pattern vector signal, The decoded vector signal pv.
．． Based on the reproduction vector 1, J torque signal v. ．． a reproducing circuit that calculates the difference vector signal 1, J and sends it to the generation circuit or display circuit,
The still image information is transmitted in stages so as to gradually become finer, and the time until the first still image is displayed is reduced by a factor of 11.

[Operation] The operation on the transmission side, which is a feature of the static 112 image transmission device according to the present invention, will be explained.

First, the device intra-frame encodes still image information input from a camera to generate index data corresponding to each block. (Step 1) Then, one frame's worth of information is transmitted to the receiving side device by using several bits (11 bits) from the upper bits of the index data corresponding to each block as transmission index data, and the transmission index data is Decode to generate a decoded vector signal. (Step 2) Next, based on the decoded vector signal generated in step 2, a reproduced vector signal corresponding to each block is calculated. (Step 3) Then, the vector signal stored in the preprocessing circuit is read a second time, and a difference vector signal between the vector signal and the reproduced vector signal group generated at the previous readout is calculated and encoded. and generate index data. (Step 4) Next, transmission index data having a longer redata length than the previous frame (by increasing 1) is generated in the same manner as in step 2 above, and is transmitted to the receiving device (step 5).

As described above, the apparatus repeats steps 2 to 5 several times.

Therefore, the still image transmission device according to the present invention applies the interframe vector quantization coding method to still image transmission, and further compresses the transmission information at the time of first vector signal readout, and is transmitted in stages, so a schematic image of the still image is displayed on the receiving side within a short time after the transmission start instruction is input.

[Embodiment] Hereinafter, a preferred embodiment of the still image transmission device of the present invention will be described based on the drawings.

FIG. 1(A) shows the block configuration of the transmitting side according to a preferred embodiment of the still image transmission device according to the present invention, and FIG. 1(B) shows the block configuration of the receiving side. There is.

In the figures, parts that are the same as those of the conventional device are designated by the same reference numerals, and their explanations will be omitted.

The feature of this embodiment is that, as shown in FIG. 1(A), in the transmitting section, the input frame memory (1a
), a differential vector signal generation circuit (10), an index data control circuit (11), a decoding circuit (12), a reproduction signal generation circuit (13), and a frame memory (14).
).

Then, the differential vector signal generation circuit (10) reads each vector signal V stored in the input frame memory (1a) for one frame every predetermined period, and reads each vector signal V at the time of initial reading.

(j=1 to n) as each differential vector signal D at the time of initial reading
V 1 , j(J-”~n), second
Difference vector signal D when reading the vector signal after the first time
Correlation between Vi,j (i=2.3...,) and the i-1 reproduced vector signal group v,, <j=1 to n) generated at 1J as described later. Benefit 1-1. For example, the difference between the current person's cover block j and the reproduced vector signal ■, , located at the same position is calculated as shown in the following equation and output.

DVi,j'''Vj-vi-1,j where i: Current number of readings'' Niblock number Also, as shown in FIG.
) and a multiplex conversion circuit (llb).

Then, the index control circuit (lla)
As shown in Figure (A), there is a frame number counter (llc) that counts the number of reads i (hereinafter referred to as frame number i), and a data length of transmission index data of 1 to 1 based on the frame number i. Length data L indicating
1, and each index data output from the main search vector quantization encoding circuit (2). ．． (j=1 to n) are respectively 1,
J The most significant bit information is below 1. until the bit is located.
Perform right shift and shift index data S1. ．． (i
- an index shift circuit (lie) that generates 1, J frame numbers (j=1 to n);

Further, the multiplex conversion circuit (llb) is shown in FIG. 3(A).
As shown in , the length data L1 and the shift index data S1 . ．． (i=frame number, j=1 to n), and a multiplexing circuit (11f) that multiplexes 1, J as shown in FIG.
Only bit information is extracted to generate transmission index data T I,j, and the length index L,
and the transmission index data Tl for one frame.
, (j-1 to n)1.J, and a serial conversion circuit (l1g) that converts the data into serial transmission data SdI and outputs the data.

Therefore, the index data control circuit (11) determines the data length J of the transmitted index data in proportion to the frame number (number of reads) i, and the tree search vector quantization initialization circuit (2) Each index data I1. 1 from the most significant bit of J. Only the information up to the bit is transmitted as index data T1. ．． 1.3.

The decoding circuit (12) then processes the shift index data S1. , and outputs a decoded vector signal pv, .

1, J At this time, the pattern vector groups of the third, sixth, and twelfth stages are stored in the decoding codebook (not shown), as shown in FIG.

Next, the reproduction circuit (13) calculates a reproduction vector signal 1J vi-1,j based on the decoded vector signals pv, .

Next, the operation of this embodiment will be explained.

First, the operation at the time of transmission when performing still image definition transmission in the third stage will be explained.

The still image transmission device inputs a still image signal (100) into the device via a camera (not shown).

Then, in the preprocessing circuit (1), the still image signal (100
) is A/D converted and then grouped into units of a plurality of pixels according to a predetermined rule and divided into n blocks.

The preprocessing circuit (1) then processes the still image signal (10
0), a vector signal V, (j=1 to n) (110) corresponding to each block j (j=1 to n) is generated and stored in the input frame memory (1a).

Next, the differential vector signal generation circuit (10) generates n vector signals (for one frame) from the input frame memory (1a) at V and (j=1 to n) at a predetermined period (1/3).
0 seconds).

Then, the differential vector signal generation circuit (10) converts each of the vector signals V, (j=1 to n) (110) into a tree search vector quantization encoding process as a differential vector signal at the time of initial reading. Output to path (2).

Then, the tree search vector quantization encoding circuit (2) determines pattern vectors PV,j (j-1
~n) from a codebook (not shown), and each selected pattern vector P Vbj (J = 1-n
) is output as index data I-(j-11, j to n) indicating the address on the codebook.

On the other hand, the frame number counter (11, C) of the index control circuit (11) counts the frame number (number of reads) i, and inputs the count value "1" into the data length table (11d).

The data length table (lid) includes the length data 10011 indicating the data length 1,=3 corresponding to the count value ``10011'' to the index shift circuit (11e).
, a multiplex conversion circuit (11b), and a decoding circuit (
12) respectively.

Next, the index shift circuit (lie) right-shifts each index data frame by frame as follows to generate shift index data.

That is, as shown in FIG. 2(B), when length data L1 "0011"-ffl, =3 is input, the index shift circuit (lie) inputs each index data 11. A right shift is performed until the most significant bit of j is located at the third bit from the lowest, and shift index data S I,j is generated.

Next, as shown in FIG. 3(B), the multiplexing circuit (11f) combines the length data Ll and one frame's worth of shift index data Sl, (j=1 to n).
is multiplexed.

1.J Also, as shown in Figure 3 (B), the serial conversion circuit (
llf) is the length data L at the beginning of each frame.
1. After that, serial data is generated to be the transmission index T I,j and transmitted to the other party's device (not shown) via the transmission buffer (3).

On the other hand, the length data L output from the data length table and the index shift circuit (lie), respectively.
1 and shift index data Sl, (j-1~
n) is also input to the decoding circuit (12) I.J.

Then, each shift index data S11. Based on j, a corresponding decoded vector signal (112) from the third stage codebook is output to the reproduction circuit (13).

Then, in the reproduction circuit (13), a reproduction vector signal is generated based on the decoded vector signal (112), and is stored in the frame memory (13b).

Therefore, each vector signal (2) of the first frame (first read) is intra-frame encoded, but the index data to be actually transmitted is greatly reduced in the index data control circuit (11). (In this embodiment, the time required to reproduce and display the first frame is reduced to approximately 1/4.) Therefore, the time required for the receiving device to reproduce and display the first frame is approximately 1/4 of the conventional time.

Next, transmission of the second frame (at the time of second readout) will be explained.

First, the differential vector signal generation circuit (10) reads the vector signal V from the input frame memory (1a), and at the same time, reproduces the reproduced vector signal of the previous frame (at the time of first reading) located at the same position as the current read block J. Read V,,j.

The differential vector signal generation circuit (10) then generates the vector signal V, the reproduced vector signal vll,
A difference vector signal DV corresponding to block j and a distortion value d2. Calculate j by 2・J and make a determination as shown in the following formula.

d,j=l DV2. When l, d,,,≧α
When d, it is invalid; when j<α, it is valid. However, α is a positive number close to 0. At this time, when the determination described in becomes invalid, the invalid information of the block is treated as the same as the information of the previous frame. only, and encoded transmission is not performed.

In addition, if the discrimination described in - is valid, the differential vector signal DV2. j is output to the main search vector quantization encoding circuit (2).

Further, the validity/invalidity information corresponding to each block j (j=1 to n) is sent to the transmission buffer (3).

The effective block is the quantized block of the first frame.
Tree search quantization is performed in the same way as during encoding, and index data 122. is generated and input to the index data control circuit (11).

Then, the frame number counter (lie) in the index control circuit (lla) outputs the counter value "2" to the data length table (lid).

Next, the data length table (lid) is a data length table corresponding to the counter value "2". =6, and outputs 2''0110''.

Then, the index shift circuit (lie) converts the index data I2. In order to use the information from the most significant bit of j to 6 bits as the transmission index data,
Each of the index data I2. A right shift is performed until the information of the most significant bit of j is located at the 6th bit from the lowest, and shift index data SI2. Generate j.

Then, similarly to the first frame, the multiplex conversion circuit (11
In b), serial index data is generated and transmitted to the other party's device (not shown) via the transmission buffer (3).

Then, in the decoding circuit (12), each block is invalidated/
After generating a decoded vector signal based on the valid information, the length data L2, and the shift index data SI9, a reproduced vector signal V2. Calculate j and store it in the frame memory.

Next, the differential vector signal Dv3.corresponding to the vector signal of the third frame (at the time of third readout). j is the same as the second frame, and the reproduced signal V2. Generate based on j.

Then, in the same manner as the second frame, the index control circuit (11) generates transmission index data with a data length of 13-12 and transmits it.

Therefore, according to this embodiment, by using intraframe encoding, at the time of initial reading (first frame) with a large amount of transmitted information, the transmission index data length is changed to I! after encoding. By making , extremely small, the amount of encoded information is significantly reduced, and the distortion value of each difference vector signal during subsequent readout (from the second frame onwards) approaches O, so the number of invalid blocks increases, so transmission Even if the index data length is increased, image transmission can be performed in a short transmission time.

Next, the operation at the time of reception will be explained.

First, encoded data is received via a communication line and stored in a reception buffer (5).

Then, the serial data of the first frame is manually input to the demultiplexing circuit (15).

Then, the received serial data is converted into length data L1.
and each transmission index data T I,j (J-1~
Separate into n) and.

Next, similarly to the decoding circuit (13) of the transmitter, the signal is decoded using the third-stage codebook to generate a reproduced signal and then displayed.

Also, when receiving the second frame and third frame data,
From the invalid/valid information and serial data of each block, length data T1, transmission index data T1.
．． The data is separated into two frames, decoded by an interframe difference of 1+J, and played back.

In the above embodiment, an example was shown in which the total number of transmission frames was 3 and each data length was 3.6.12, but the same effect as in the above embodiment can be obtained even if the total number of transmission frames and data length are arbitrary. play.

In addition, in the second embodiment, an example was shown in which the length data is added and transmitted, but if the total number of transmission frames and the corresponding length data I7 are agreed upon in advance on the transmitting and receiving sides, the length data can be added. Even without the transmission of , the same effects as in the above embodiment can be achieved.

Further, in the above embodiment, an example using binary search vector quantization was shown, but even if the index data is an address on a codebook or a pattern vector arranged in a multi-ary tree structure, It has the same effect as.

``Effects of the Invention'' As described in Section 2, according to the present invention, it is possible to reduce the playback time of the first stage image data from the start of still image transmission to I11, and the image content can be quickly conveyed to the user. It has the effect of being able to

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a transmitting section and a receiving section of a preferred embodiment of a still image transmission device according to the present invention, FIG. 2 is a detailed explanatory diagram of an index control circuit according to the embodiment of FIG. 1, and FIG.
1 is a detailed explanatory diagram of the multiplexing conversion circuit according to the embodiment in FIG. 1, FIG. 4 is an explanatory diagram of the binary tree structure of the codebook according to the embodiment in FIG. 1, and FIG. 5 is a decoding diagram according to the embodiment in FIG. 1. FIG. 6 is an explanatory diagram of the multiplexing and demultiplexing circuit according to the embodiment of FIG. 1, FIG. 7 is an explanatory diagram of the conventional image encoding and transmitting device,
FIG. 8 is an explanatory diagram of a conventional search vector quantization unit, and FIGS. 9 and 10 are explanatory diagrams of a codebook. In the figure, (2) is the main search vector quantization unit, (3)
is a transmission buffer, (4a) and (4b) are communication lines,
(5) is a reception buffer, (6) is a decoding unit, (10) is a differential vector signal generation circuit, (11) is an index data control circuit, (lla) is an index control circuit, (
llb) is a multiplex conversion circuit, (11c) is a frame number counter, (11d) is a data length table, (lie) is an index shift circuit, (11t) is a multiplexing circuit, (
l1g) is a serial conversion circuit, (12) a decoding circuit, (13)
) is a frame memory, and (15) is a demultiplexing circuit. In addition, the same No. 71 is shown in the figure, or a corresponding part.

Claims (1)

[Claims] A still image transmission device that compresses and encodes a still image signal and transmits and receives the still image signal, reads the still image signal, A/D converts the image signal to generate a pixel signal, and converts the pixel signal into a predetermined signal. A pre-processing circuit that generates n blocks by combining multiple blocks according to a rule, generates a vector signal V_j (where j indicates a block number) corresponding to each block j, and stores each vector signal; Each time, n vector signals (one frame worth) stored in the preprocessing circuit are repeatedly read, and a differential vector signal DV_i_, corresponding to each vector signal V_j is generated.
It is a circuit that outputs each differential vector signal DV_i_ at the time of initial reading.
, _j (j=1 to n), each of the vector signals V_j
(j=1 to n), and output the differential vector signal DV_i_,_j (i=2, 3, ..., j=1 to n) during subsequent reading.
) is the reproduced signal group when last read ■_i_-_1_,_j
(j=1 to n); a codebook having pattern vectors arranged in a tree structure; and a codebook having pattern vectors arranged in a tree structure; Input the difference vector signals DV_i_,_j, select the pattern vector signal most similar to the difference vector signals DV_i_,_j from the codebook, and provide index data I_i_, indicating the address on the codebook of the selected pattern vector. A tree search vector quantization encoding circuit that outputs _j, and a tree search vector quantization encoding circuit that outputs transmission index data TI_i_ that is actually transmitted according to the number of readings i, as the number of readings i increases.
, _j such that the number of bits l_i of the index data I_i_, _j is increased, and data from the most significant bit (first stage side of tree search) to the l_i-th bit of each index data I_i_, _j is output as transmission index data TI_i_, _j. a data control circuit; a transmission control circuit that transmits or receives the transmission index data TI_i_, _j for each frame for each reading count; and the transmitted or received transmission index data TI_i.
A pattern vector signal corresponding to
a decoding circuit that generates _, _j, and a reproduction vector signal ■_ based on the decoded vector signals PV_i_, _j;
A still image transmission device comprising: a reproduction circuit that calculates i_, _j and sends it to the difference vector signal generation circuit or display circuit.