JPH0965337A - Image data processing apparatus and image data processing method - Google Patents

Abstract

(57) Abstract: Image data is compressed at a high compression rate. In a coding device, a zero region prediction device (101) that predicts a zero region of a high layer from a low layer, and a zero region mask device (102) that deletes image data predicted to be unimportant out of data after wavelet transform. In addition, the decoding device further includes a zero region prediction device that predicts a zero value region of a high layer from a low layer, and an inverse device that restores the data after wavelet transform from the predicted data in which the unimportant part is deleted. Equipped with a wavelet transform device.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data processing device and an image data processing method used for image compression processing and the like.

[0002]

2. Description of the Related Art DCT (Discrete Cosine Transform) is used in high-efficiency image coding technology for communication media and recording media.
The technology using is widely applied. However,
As an intrinsic problem inherent in the compression method using DCT,
When the compression rate was increased, block distortion, mosquito noise, etc. were visually recognized, and the compression rate was limited.

Therefore, in recent years, a new compression method has been proposed in order to improve the compression rate, and in particular, a compression technique using a wavelet transform which is one of subband coding (hereinafter referred to as a wavelet). Attention has been paid. When this wavelet is used, there is no concept of block, so that the interblock distortion generated in DCT is eliminated, and a considerable improvement in image quality can be expected visually.

A feature of compressed images of DCT and wavelet is that the DCT becomes a digital image, and by increasing the compression rate, it becomes an image composed of high frequency components. As a result, the high-frequency component is preserved, but on the contrary, distortion composed of visually conspicuous high-frequency components occurs. On the other hand, the wavelet becomes an analog image, and the high frequency component is naturally lost by increasing the compression rate. In other words, it is gradually cut from the high frequency part of the signal band. As a result, the resolution is reduced as a whole, so that the visual quality degradation is reduced as compared with DCT at the same compression ratio.

A conventional wavelet compression device will be described below.

FIG. 10 is a block diagram showing the overall structure of a conventional wavelet image compression apparatus. 1001 is an input image,
1002 is a wavelet transform device, 1003 is a code quantization device, 1004 is a variable length coding device (VLC), 1005 is a variable length decoding device (IVLC), 1006 is an inverse code quantization device, 1007 is an inverse wavelet transform device, 1008 is a decoded image, 1021 is input image data, 1022 is layer 1 (L1.) Subband data, 102
3 Layer 2 (L2.) Subband data, 1024 is Layer 3
(L3.) Subband data, 1025 is layer 1 quantized subband data, 1026 is layer 2 quantized subband data, 1027 is layer 3 quantized subband data, 1028 is variable length compressed image data, and 1029 is decoded. Layer 1 quantized subband data, 1030 decoded layer 2 quantized subband data, 1031 decoded layer 3 quantized subband data, 1032 decoded layer 1 subband data, 1033 decoded Layer 3 subband data, 1034 is decoded layer 3 subband data, and 1035 is decoded image data.

FIG. 11 is a block diagram showing the structure of the wavelet transform device. 1101 is a two-dimensional filter bank, 1121 is layer 1 LL component subband data, 11
22 is the layer 1 LH component subband data, 1123 is the layer 1 HL component subband data, 1124 is the layer 1 HH component subband data, and 1125 is the layer 2 L
H component subband data, 1126 is layer 2 HL component subband data, 1127 is layer 2 HH component subband data, 1128 is layer 3 LH component subband data, 1129 is layer 3 HL component subband data,
1130 is HH component subband data of layer 3, 1131 is LL component subband data of layer 3, and 1132 is LL component subband data of layer 2.

FIG. 12 is a block diagram showing the structure of the two-dimensional filter bank shown in FIG. 11. 1201 is a horizontal low-pass filter, 1202 is a horizontal high-pass filter, 1203 is a vertical low-pass filter, 1204 is a vertical high-pass filter, and 1205 is a horizontal down filter. Sampler, 1206 vertical down sampler, 1221 two-dimensional data, 1222 horizontal component low frequency sub-band data, 12
23 is horizontal component high frequency subband data, 1224 is LL component subband data, 1225 is LH component subband data,
1226 is HL component subband data, and 1227 is HH component subband data.

FIG. 13 is an explanatory view of the filter characteristic in the two-dimensional filter bank, where 1301 is the frequency characteristic of the horizontal / vertical low pass filter and 1302 is the frequency characteristic of the horizontal / vertical high pass filter.

FIG. 14 is an explanatory diagram of the down sampler operation. 1401 is a signal sample interval, 1402 is a sub sample interval, 1403 is a first sample, 1404 is a second sample, and 14 is a second sample.
05 is the third sample and 1406 is the fourth sample.

FIG. 15 is an explanatory diagram of the wavelet space, and 1501 is the LL component subband space of layer 1, 1502.
Is the HL component subband space of layer 1, 1503 is the LH component subband space of layer 1, and 1504 is the H of layer 1
H component subband space, 1505 layer 2 HL component subband space, 1506 layer 2 LH component subband space, 1507 layer 2 HH component subband space, 15
08 is the HL component subband space of layer 3, 1509 is the LH component subband space of layer 3, and 1510 is the HH component subband space of layer 3.

FIG. 16 is a block diagram showing the structure of the code quantizer of FIG. 10. 1601 is an integer divider, 1602 is a near-zero clipping device, 1621 is LL component quantized subband data of layer 1, and 1622 is Layer 1 of clip unprocessed layer 1
H component quantized sub-band data, 1623 is the HL component quantized sub-band data of layer 1 that has not been clipped, 1624
Is the HH component quantized sub-band data of layer 1 which is not clipped, 1625 is the LH of layer 2 which is not clipped
Component quantized sub-band data, 1626 HL component quantized sub-band data of clip-unprocessed layer 2, 1627 is HH component quantized sub-band data of clip-unprocessed layer 2, 1628 is clip-unprocessed layer 3 LH component quantized sub-band data, 1629 is HL component quantized sub-band data of unclip-processed layer 3, 1630 is HH component quantized sub-band data of clip-unprocessed layer 3, 1631 is LH component quantized of layer 1. 1632 is the layer 1 HL component quantized subband data, 1633 is the layer 1 HH component quantized subband data, 1634 is the layer 2 LH component quantized subband data, and 1635 is the layer 2 HL component. Quantized subband data, 1636 is HH component of layer 2 Quantized subband data, 1637 is LH component of layer 3 Coca sub-band data, 1638 HL component quantized sub-band data of the layer 3, 1639 is a HH component quantized sub-band data of the layer 3.

FIG. 17 is a diagram showing the structure of the VLC of FIG. 10, in which 1701 is a PS converter, 1702 is a lossless encoder, and 17
Reference numeral 21 is compressed continuous data.

FIG. 18 is a diagram showing the structure of the IVLC of FIG. 10, in which 1801 is a lossless decoding device, 1802 is a P-S conversion device, and 18
21 is the expanded continuous data, 1822 is the decoded layer 1 L
L component quantized subband data, 1823 is the decoded layer 1 LH component quantized subband data, 1824 is the decoded layer 1 HL component quantized subband data,
1825 is the decoded layer 1 HH component quantized subband data, 1826 is the decoded layer 2 LH component quantized subband data, and 1827 is the decoded layer 2 H
L component quantized subband data, 1828 is decoded layer 2 HH component quantized subband data, 1829 is decoded layer 3 LH component quantized subband data,
1830 is the decoded HL component quantized subband data of layer 3, and 1831 is the decoded HH component quantized subband data of layer 3.

FIG. 19 is a block diagram showing the structure of the inverse code quantizer of FIG. 10. 1901 is a multiplier, 1921 is decoded layer 1 LL component band data, and 1922 is decoded layer 1 LH. Component band data, 1923 is the decoded layer 1 HL component band data, 1924 is the decoded layer 1 HH component band data, 1925 is the decoded layer 2 LH component band data, and 1926 is the decoded layer 2 1927 is the decoded layer 2 HH component band data, 1928 is the decoded layer 3 LH component band data, 1929 is the decoded layer 3 HL component band data, and 1930 is the decoded It is HH component band data of layer 3.

FIG. 20 is a block diagram showing the structure of the inverse wavelet transform device shown in FIG. 10, where 2001 is an inverse two-dimensional filter bank, 2021 is the decoded LL component subband data of layer 2, and 2022 is the decoded layer. 3 is the LL component subband data of No. 3.

FIG. 21 is a block diagram showing the structure of the inverse two-dimensional filter bank, 2101 is a band synthesis filter based on a horizontal low pass filter, 2102 is a band synthesis filter based on a horizontal high pass filter, and 2103 is a vertical low pass filter. 2104 is a horizontal upsampler, 2106 is a vertical upsampler, and 2121 is L.
L component subband data, 2122 LH component subband data, 2123 HL component subband data, 2124 HH component subband data, 2125 horizontal component low frequency subband data, 2126 horizontal component high frequency subband data, 2127
Is two-dimensional data.

FIG. 22 is an explanatory diagram of the operation of the upsampler. 2201 is a signal sample interval, 2202 is an upsample interval, 2203 is a first sample, 2204 is a second sample, 22.
05 is the first zero insertion and 2206 is the second zero insertion.

Next, the operation of the conventional device will be described.

First, if an input image 1001 is given,
The wavelet transform device 1002 receives the input image data 1021 and uses the two-dimensional filter bank 1101 of the wavelet transform device 1002 of FIG.
22, layer 2 subband data 1023, and layer 3 subband data 1024 are generated. The detailed operation of the wavelet transform device 1002 will be described here. In the first two-dimensional filter bank 1101 to which the input image data 1021 is input, the two-dimensional data 1221 shown in FIG. Then, horizontal component low-frequency subband data 1222 and horizontal component high-frequency subband data 1223 are generated. This is further sub-band-divided in the vertical direction and 4
LL component subband data 1224 of the seed subband component,
The LH component subband data 1225, the HL component subband data 1226, and the HH component subband data 1227 are generated.

LH component subband data obtained here
1225, HL component subband data 1226, and HH component subband data 1227 become layer 3 wavelet component data. Further, the LL component subband data 1131 of layer 2 obtained here is further generated by the second two-dimensional filter bank 1101 to newly generate wavelet component data of layer 2, and the LL component subband data of layer 2 obtained at the same time. 1132 is further generated by the third two-dimensional filter bank 1101 to newly generate layer 2 wavelet component data, and the layer 2 LL component subband data 1132 obtained at the same time is also used as wavelet component data.

A wavelet space shown in FIG. 15 is obtained by arranging wavelet component data obtained by the wavelet transform device 1002 based on frequency components. Here, a wavelet component having a high horizontal frequency component from the upper left to the right of the space. Data is arranged, and wavelet component data having a high vertical frequency component is arranged from the upper left of the space to the left. As far as the wavelet space is viewed, the amount of data forming an image is the same as the amount of data of the input image 1001.

Here, the wavelet component data is irreversibly compressed from the visual characteristics by the code quantizer 1003. The code quantizing device 1003 is composed of a plurality of integer dividing devices 1601 and a plurality of near-zero clipping devices 1602 as shown in FIG. 16, and the integer dividing device 1601 performs degeneration of the dynamic range, and near-zero clipping devices 1602. Then, delete the less important information. Further, to each subband data, a degeneration parameter in the integer division device 1601 and a clip parameter in the near zero cribping device 1602 are given to each subband data from each characteristic. The compression ratio and image quality are affected by the way this parameter is given.

In the wavelet component data, the more important the information, the lower the frequency component. Conversely speaking, the sub-band data of the high frequency component has little deterioration of the decoded image even if it is largely lost. In the wavelet image compression device, paying attention to this, the dynamic range of the layer 1 quantized subband data 1025, the layer 2 quantized subband data 1026, and the HH component subband data 1127 of the layer 2 containing a lot of high frequency components is reduced. , Clipped unprocessed layer 2 LH component quantized subband data 16
25, HL component quantized subband data 1626 of clip-unprocessed layer 2, HH of clip-unprocessed layer 2
Layer 2 suitable for lossless data compression including run lengths by generating component quantized sub-band data 1627 and reducing all the data in the vicinity of zero to further reduce the amount of information.
LH component quantized subband data 1634, layer 2 HL component quantized subband data 1635, layer 2 H
The H component quantized subband data 1636 is generated.

Layer 1 quantized subband data 1025, layer 2 quantized subband data 1026, and layer 3 quantized subband data 1027 that have been lossy compressed by the encoding / quantizing device 1003 are integrated to further change lossless compression. This is performed by the long encoding device 1004. In the variable length coding device 1004, FIG.
As shown in FIG. 2, each subband data is concatenated by the PS converter 1701 so that the 0 run length becomes long, and compressed continuous data 1721 is generated. Reversible compression such as encoding is performed to obtain variable length compressed image data 1028.

Further, in order to decode the variable length compressed image data 1028, as shown in FIG. 18, the variable length decoding device 10 is used.
The lossless code decoder 1801 of 05 performs decoding processing based on the lossless compression process to generate expanded continuous data 1821, and
In the P conversion device 1802, each decoded layer 1 LL component quantized subband data 1822 and decoded layer 1 LH component quantized subband data 1823 are decoded based on the subband data concatenation process. Layer 1
HL component quantized sub-band data 1824 and decoded layer 1 HH component quantized sub-band data 1825,
The decoded layer 2 LH component quantized subband data 1826, the decoded layer 2 HL component quantized subband data 1827, the decoded layer 2 HH component quantized subband data 1828, and Layer 3
LH component quantized sub-band data 1829, decoded layer 3 HL component quantized sub-band data 1830,
The decoded HH component quantized subband data 1831 of layer 3 is reproduced.

In response to this, the inverse code quantizer 1006 performs multiplication based on the dynamic range degeneracy division parameter in the multiplier 1901 and decodes the decoded layer 1 LL.
Component subband data 1921, decoded layer 1 L
H component subband data 1922, decoded layer 1 HL component subband data 1923, decoded layer 1
HH component subband data 1924, decoded layer 2 LH component subband data 1925, decoded layer 2 HL component subband data 1926, decoded layer 2 HH component subband data 1927, decoded LH component subband data 1928 of layer 3, decoded HL component subband data 1929 of layer 3, and decoded HH component subband data 1930 of layer 3 are generated. An inverse wavelet transform device 1007 obtains a decoded image 1008 based on the generated subband data.

[0028]

As described above, in the above-mentioned conventional apparatus, the image data is compressed by using the wavelet transform, but in the lossy compression process, the information having a zero value is biased and compressed. I could not expect a high compression rate.

It is an object of the present invention to provide an image data processing device and an image data processing method which solve the above-mentioned conventional problems and can achieve a high compression rate.

[0030]

According to the present invention, for encoding, a means for predicting a zero value region from a low layer to a high layer, and image data predicted to be insignificant among data after wavelet transform are encoded. Equipped with a means for deleting, and for decoding, a means for predicting the zero value area of the high layer from the low layer, and a method for recovering the wavelet transformed data from the predicted image data in which the unimportant part has been deleted. Means for doing so.

[0031]

According to the present invention, by providing means for predicting a zero-value area from a low layer to a high layer in encoding and decoding, the data in the area predicted to be the zero-value area can be reduced from the compressed data. High data compression can be realized.

[0032]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an image data processing method of the present invention and an image data processing apparatus for implementing the method will be described below with reference to the drawings. Members and the like corresponding to the members and the like described with reference to FIGS. 10 to 22 are denoted by the same reference numerals and detailed description thereof will be omitted.

FIG. 1 is a block diagram showing the overall structure of a wavelet image compression apparatus according to the first embodiment of the present invention. 101 is a zero area prediction apparatus, 102 is a zero area mask apparatus, 12
Reference numeral 1 is layer 2 zero area prediction data, 122 is layer 3 zero area prediction data, 123 is layer 2 quantization mask subband data, and 124 is layer 3 quantization mask subband data.

FIG. 2 is a block diagram showing the structure of the wavelet image decompression device in the first embodiment. 201 is a zero area interpolating device, 221 is decoded layer 2 quantization mask subband data, and 222 is decoded. Layer 3 quantization mask subband data, 223 is decoded layer 2 zero area prediction data, 224 is decoded layer 3 zero area prediction data, 225 is decoded layer 2 mask subband data, and 226 is decoded Layer 3 mask subband data.

FIG. 3 is a block diagram showing the structure of the zero region prediction apparatus of FIG. 1, where 301 is a central processing unit (CPU) for zero region prediction, 302 is a subband data input port, and 303 is prediction data output. Port, 304 is mask plane memory, 3
Reference numeral 05 is a working memory for zero area prediction, and 321 is a system bus for zero area prediction.

FIG. 4 is a flow chart relating to the zero area prediction processing in this embodiment (detailed later).

FIG. 5A is an explanatory view showing the process of releasing the mask plane, FIG. 5B is an explanatory view of the transfer data, 501 is a mask plane space corresponding to a wavelet space,
Hereinafter, with reference to FIG. 15, 502 is a mask plane space of the LL component subband space 1501 of layer 1, and 50
3 is a mask plane space of the HL component subband space 1502 of layer 1 (hereinafter referred to as mask plane space 503), 5
04 is the mask plane space of the LH component subband space 1503 of layer 1, 505 is the mask plane space of the HH component subband space 1504 of layer 1, and 506 is the H of layer 2
Mask plane space of L component subband space 1505 (hereinafter referred to as mask plane space 506), 507 is layer 2
LH component sub-band space 1506 of the mask plane space,
508 is the mask plane space of the HH component subband space 1507 of layer 2, 509 is the mask plane space of the HL component subband space 1508 of layer 3 (hereinafter referred to as mask plane space 509), and 510 is the LH component sub of the layer 3. Band plane 1509 mask plane space, 511 is layer 3
HH component sub-band space 1510 mask plane space,
Reference numeral 521 is transfer target prediction source data in the mask plane space 503, 522 is transfer target data in the mask plane space 506, and 523 is transfer target data in the mask plane space 509.

In FIG. 5B, 551 is the first transfer data in the mask plane space 506, 552 is the second transfer data in the mask plane space 506, and 553 is the third transfer data in the mask plane space 506. Data, 554 is the fourth transfer data in the mask plane space 506, 555 is the first transfer data in the mask plane space 509, 5
56 is second transfer data in the mask plane space 509, 557 is third transfer data in the mask plane space 509, 558 is fourth transfer data in the mask plane space 509, and 559 is in the mask plane space 509. Fifth transfer data, 560 is sixth transfer data in the mask plane space 509, 561 is seventh transfer data in the mask plane space 509, and 562 is the mask plane space 509.
In the mask plane space 509, 564 is the 10th transfer data in the mask plane space 509, 565 is the 11th transfer data in the mask plane space 509, 566 is the twelfth transfer data in the mask plane space 509.

FIG. 6 is a block diagram showing the structure of the zero area mask device of FIG. 1, 601 is a central processing unit (CPU) for the zero area mask, 602 is a prediction / subband input port, and 60
3 is a mask processing sub-band data output port, 604 is a transfer buffer memory, 605 is a working memory for the zero area mask, and 621 is a system bus for the zero area mask.

FIG. 7 is a flow chart of the zero area mask processing in this embodiment (detailed later).

FIG. 8 is a block diagram showing the structure of the zero area interpolation device of FIG. 2, in which 801 is a central processing unit (CPU) for zero area interpolation, 802 is a mask / subband data input port, and 803 is interpolation. Processing subband data output port, 804 is wavelet spatial memory, 805 is working memory, 821
Is a system bus for zero area interpolation.

FIG. 9 is a flow chart of the zero area interpolation processing in this embodiment (detailed later).

Next, the operation of this embodiment will be described.

First, if an input image 1001 is given,
The wavelet transform device 1002 receives the input image data 1021, and uses the two-dimensional filter bank 1101 (see FIG. 11) of the wavelet transform device 1002 to generate layer 1 subband data 1022, layer 2 subband data 1023, and layer 3
Generate subband data 1024.

The layer 2 sub-band data 1023 is irreversibly compressed based on the layer 1 sub-band data 1022, the layer 2 sub-band data 1023 and the layer 3 sub-band data 1024 obtained by the wavelet transform device 1002. 1-quantized subband data 1025, layer 2 quantized subband data 1026, and layer 3 quantized subband data 1027 are generated.

The zero area prediction device 101 predicts the position of important data of the layer 2 quantized subband data 1026 based on the layer 1 quantized subband data 1025, and the layer 2 zero area predicted data indicating the predicted position. Generate 121. Similarly, layer 2 quantized subband data 1026
The position of the important data of the layer 3 quantized sub-band data 1027 is predicted based on the
The zero area prediction data 122 is generated.

The prediction method of the zero region prediction apparatus 101 will now be described with reference to FIG. 15 already described. When predicting the position of important data of the layer 2 quantized subband data 1026, Note that there is a correlation between the HL subband space 1502, the layer 2 HL subband space 1505, and the layer 3 HL subband space 1508, and is larger than an arbitrary threshold value of the layer 1 HL subband space 1502. By doubling the position of a component having a value in both the horizontal and vertical directions from the component having a value, the position where the component having a large value in the HL subband space 1505 of layer 2 is predicted, and the predicted position is calculated. To generate a mask plane space 506. Then, only this predicted region is the target of compressed image data. Similarly, only the region predicted to have a large value of the layer 2 HL sub-band space 1505 has the same layer 3 HL.
A position where a component having a large value in the subband space 1508 exists is predicted, and a mask plane space 509 representing this predicted position is generated.

In the zero region prediction apparatus 101, the above operation is performed on the HH subband space 1510 of layer 1 and the mask plane space of the HH subband space 1507 of layer 2 is performed.
508 and the mask plane space 511 of the HH subband space 1510 of the layer 3 are generated, and the layer 2 is generated for the LH subband space 1503 of the layer 1, the LH subband space 1506 of the layer 2 and the LH subband space 1509 of the layer 3. A mask plane space 507 of the LH sub-band space 1506 and a mask plane space 510 of the layer 3 LH sub-band space 1509 are generated.

For example, suppose that the transfer target prediction source data 521 in the mask plane space 503 has a value equal to or greater than a preset threshold value as a parameter.
The horizontal position and the vertical position of the transfer target prediction source data 521 in the mask plane space 503 are each doubled to obtain the coordinates of the transfer target region of the next layer, and from that position, the mask plane space 506 having an area four times larger. First in
Of the transfer data 551, the second transfer data 552 in the mask plane space 506, the third transfer data 553 in the mask plane space 506, and the fourth transfer data 554 in the mask plane space 506 of the layer 2 A mask plane space 506 is generated as a transfer target area of the HL subband space 1505.

Assuming that the mask plane space 506 is
First transfer data 551 in the mask plane space 506
Suppose that only the second transfer data 552 in the mask plane space 506, the third transfer data 553 in the mask plane space 506, and the fourth transfer data 554 in the mask plane space 506 are transfer targets. , The zero region prediction apparatus 101 uses the HL sub-band space 150 of layer 3.
Since the eight transfer targets are propagated from the mask plane space 506, the first transfer data 551 in the mask plane space 506 and the first transfer data 551 in the mask plane space 506 among the transfer data 522 in the mask plane space 506. Assuming that the second transfer data 552 and the third transfer data 553 in the mask plane space 506 have a value equal to or larger than a preset threshold parameter, the first transfer data 551 in the mask plane space 506 becomes On the other hand, the first transfer data 555 in the mask plane space 509
, The second transfer data 556 in the mask plane space 509, the fifth transfer data 559 in the mask plane space 509, and the sixth transfer data 560 in the mask plane space 509 are transfer targets. Five
For the second transfer data 552 in 06, the third transfer data 557 in the mask plane space 509,
Fourth transfer data 558 in mask plane space 509
And the seventh transfer data 561 in the mask plane space 509 and the eighth transfer data 562 in the mask plane space 509 are transfer targets, and the third transfer data 553 in the mask plane space 506 is masked with respect to the third transfer data 553. Ninth transfer data 563 in the plane space 509, tenth transfer data 564 in the mask plane space 509, eleventh transfer data 565 in the mask plane space 509,
12th transfer data 566 in the mask plane space 509
Becomes a transfer target, and a mask plane space 509 is generated.

As described above, the zero area prediction apparatus 101
The obtained mask plane space 506 and LH of layer 2
Mask plane space 507 of subband space 1506 and mask plane space of HH subband space 1507 of layer 2
The mask information obtained in 508 is used as the layer 2 zero area prediction data 121, and the mask plane space 509 and the mask plane space 51 of the LH subband space 1509 of layer 3 are used.
0 and the mask information obtained in the mask plane space 511 of the HH subband space 1510 of layer 3 are generated as layer 3 zero area prediction data 122.

The operation of the zero area predicting device 101 will be described with reference to the flowchart of FIG. 4. First, in step 401 for predicting the zero area, the front area of the mask plane is masked. Step 402 for zero area prediction
At, it is determined whether or not the prediction of layer 1 is completed. In step 403 for zero region prediction,
When the first layer 1 quantized sub-band data 1025 is input and the input data is equal to or less than the arbitrarily set threshold value in step 404 for zero area prediction, the step 402 for zero area prediction is performed again. Layer 1 in step 405 for zero region prediction if greater than threshold
The mask coordinates of layer 2 are calculated from the input coordinates of the quantized sub-band data 1025, and the mask plane is partially unmasked based on the coordinates calculated in step 406 for the zero area prediction, and the zero area prediction is performed. Steps for
Return to 402.

When the prediction of the layer 1 is completed for all of the layer 1 quantized subband data 1025, the step 402 for the zero area prediction shifts to the step 407 for the zero area prediction. In step 407 for zero area prediction, it is determined whether or not prediction of the final layer, Layer 2, has been completed. Step 40 for null region prediction
First layer 2 quantized subband data 1 in 8
If 026 is input and the input data in step 409 for the zero region prediction is equal to or less than the arbitrarily set threshold value, the process proceeds to step 407 for the zero region prediction again, and if it is larger than the threshold value, the zero region is input. In step 410 for prediction, the mask coordinates of layer 3 are calculated from the input coordinates of the layer 2 quantized sub-band data 1026, and the partial mask plane is calculated based on the coordinates calculated in step 411 for zero region prediction. The mask is released, and the process returns to step 407 for zero area prediction. When the prediction of layer 2 is completed for all the layer 2 quantized subband data 1026, the prediction process ends.

The zero area mask device 102 masks from the layer 2 quantized subband data 1026 and the layer 3 quantized subband data 1027 based on the layer 2 zero area predicted data 121 and the layer 3 zero area predicted data 122. The layer 2 quantization mask subband data 123 and the layer 3 quantization mask subband data 124, which are composed only of valid data that have not been processed, are generated.

The operation of the zero area masking device 102 will be described with reference to the flowchart of FIG. 7. First, in step 701 for zero area masking, all data in the transfer buffer memory 604 is invalidated. In step 702 for zero area mask, it is determined whether or not the mask processing of layer 2 is completed. In step 703 for the zero area mask, the first layer 2 zero area prediction data 1
Input 21 and input the layer 2 zero area prediction data 121 input in step 704 for zero area mask to the non-transfer target area, that is, if masked, input again to step 702 for zero area mask. If the layer 2 zero area prediction data 121 thus obtained is the area to be transferred, that is, if it is not masked, the layer 2 quantized subband data 1 in step 705 for zero area masking.
026 is obtained, the layer 2 quantized sub-band data 1026 obtained in step 706 for zero area mask is written in the transfer buffer memory 604 as compressed data, and the procedure returns to step 702 for zero area mask.

When the masking processing for layer 2 is completed, the process proceeds from step 702 for zero area mask to step 707 for zero area mask. In step 707 for the zero area mask, it is determined whether or not the masking process for the final layer, Layer 3, is completed. The first layer 3 zero area prediction data 122 is input in step 708 for the zero area mask, and the layer 3 zero area prediction data 1 input in step 709 for the zero area mask is input.
If 22 is a non-transfer target area, that is, if it is masked, go to step 707 for masking the zero area again, and if the input layer 3 zero area prediction data 122 is a transfer target area, that is, if it is not masked. The layer 3 quantized sub-band data 1027 is obtained in step 710 for the zero area mask, and the layer 3 quantized sub-band data 1027 obtained in step 711 for the zero area mask is used as compressed data in the transfer buffer memory 60.
Write to 4 and return to step 707 for zero area mask. When the masking processing of layer 3 is completed, the masking processing ends.

The variable length coding device 1004 integrates the layer 1 quantized subband data 1025, the layer 2 quantized mask subband data 123 and the layer 3 quantized mask subband data 124, and further performs lossless compression, Variable length compressed image data 1028 which is compressed data is generated.

Further, in order to decode the variable length compressed image data 1028, the variable length decoding device 1005 performs a decoding process based on the variable length compressed image data 1028 based on the variable length encoding device 1004, Decoded layer 1 quantized subband data 1029, decoded layer 2 mask subband data 225 and decoded layer 3 mask subband data 226 are generated.

The inverse code quantizer 1006 of FIG. 2 is based on the decoded layer 1 quantized subband data 1029, the decoded layer 2 mask subband data 225 and the decoded layer 3 mask subband data 226. The decoded layer 1 subband data 1032, the decoded layer 2 quantization mask subband data 221, and the decoded layer 3 quantization mask subband data 222 are generated.

The zero area prediction device 101 first generates the decoded layer 2 zero area prediction data 224 based on the decoded layer 1 subband data 1032. Next, based on the decoded layer 2 mask subband data 225 generated by the zero area interpolation device 201 described later, the decoded layer 3 zero area prediction data 223 is generated.

The zero area interpolating device 201 generates the decoded layer 2 sub-band data 1033 based on the decoded layer 2 zero area prediction data 224 and the decoded layer 2 quantization mask sub-band data 221, and Decoded layer 3 sub-band data 1034 is generated based on the layer 3 zero area prediction data 223 and the decoded layer 3 quantization mask sub-band data 222.

The operation of the zero area interpolation device 201 will be described with reference to the flowchart of FIG. 9. First, in step 901 for zero area interpolation, the contents of the wavelet space memory 804 are all set to 0, which is the default value.
In step 902 for zero area interpolation, it is determined whether or not the layer interpolation processing is completed. In step 903 for zero area interpolation, the first decoded layer 2 zero area prediction data 224 is input and the decoded layer 2 input in step 904 for zero area interpolation is input.
If the zero region prediction data 224 is a non-transfer target region, that is, if it is masked, the step for zero region interpolation is performed again.
If the input decoded layer 2 zero area prediction data 224 is not the transfer target area, that is, is not masked to 902, in step 905 for zero area interpolation,
Decoded layer 2 quantization mask subband data 22
1 is obtained, and the decoded layer 2 quantization mask subband data 221 obtained in step 906 for zero area interpolation is written in the wavelet spatial memory 804 as wavelet decoded data for interpolation of zero area interpolation. Return to step 902.

When all the interpolation processing of layer 2 is completed, step 902 for zero area interpolation transits to step 907 for zero area interpolation. Step 9 for zero area interpolation
At 07, it is determined whether or not the interpolation processing for the final layer, Layer 3, is completed. Steps for zero area interpolation
In 908, the first decoded layer 3 zero area prediction data 223 is input, and the decoded layer 3 zero area prediction data 223 input in step 909 for zero area interpolation is not transferred or is masked. If the input decoded layer 3 zero area prediction data 223 is the transfer target area, that is, if the input decoded layer 3 zero area prediction data 223 is present, the processing proceeds to step 907 for zero area interpolation again. The decoded layer 3 quantization mask subband data 222 is acquired, and the decoded layer 3 quantization mask subband data 222 acquired in step 911 for zero area interpolation is used as wavelet decoded data in a wavelet space memory. Write to 804 and return to step 907 for zero area interpolation. The masking process ends when the interpolation process for Layer 3 is completed.

The inverse wavelet transform device 1007 generates decoded image data 1035 based on the decoded layer 2 subband data 1033, the decoded layer 3 subband data 1034 and the inverse code quantization device 1006, and the decoded image 1008 is generated. To get

As described above, when an element in the low frequency subband space is larger than an arbitrary value, the element exists as a concrete method for predicting it as an effective element in the next high frequency subband space. Coordinates obtained by doubling the coordinates horizontally and doubling vertically and doubling the coordinates horizontally and vertically
Coordinates obtained by (2 times +1), horizontally (2 times +1) and vertically obtained by doubling coordinates, and horizontally (2 times +1)
In addition, it is considered as an example to predict a total of 4 element coordinates of coordinates obtained by vertically (2 times +1) as an effective element of the next high-frequency subband space. To the coordinates obtained by doubling horizontally and doubling vertically, and the coordinates obtained by doubling horizontally and (2 times +1) vertically, and doubling horizontally (2 times +1) and vertically And (2 times +1) horizontally and vertically (2
Coordinates obtained by multiplying (+1 times), coordinates obtained by multiplying (2 times-1) horizontally and twice by vertically, making coordinates (2 times -1) horizontally and (2 times +1) vertically With the obtained coordinates, horizontally (2
(Double +2) and the coordinates obtained by doubling vertically and horizontally (2
(2 times +2) and vertically (2 times +1) to obtain the total 8 element coordinates as the effective element of the next high-frequency subband space, or to horizontally determine the existing coordinate of the element. Coordinates obtained by doubling and doubling vertically,
Coordinates obtained by doubling horizontally and vertically (2 times +1), and coordinates obtained horizontally (2 times +1) and vertically doubling, horizontally (2 times +1) and vertically Coordinates obtained by multiplying (2 times + 1), coordinates obtained by doubling horizontally and (2 times -1) vertically, and obtaining by doubling horizontally and (2 times + 2) vertically Of the coordinates obtained horizontally (2 times +1) and vertically (2 times -1), and coordinates obtained horizontally (2 times +1) and vertically (2 times +2) Predicting a total of eight element coordinates as valid elements in the next high-frequency subband space, or doubling the coordinates in which the elements exist horizontally and vertically, and twice the coordinates obtained horizontally. And the coordinates obtained by vertically (twice +1), the coordinates obtained horizontally (twice + 1) and vertically doubled, and the horizontal (2
And the coordinates obtained by vertically multiplying (2 times +1) and
Coordinates obtained horizontally (2 times -1) and vertically (2 times -1), and coordinates obtained horizontally (2 times +2) and vertically (2 times -1) (2 times -1) and vertically (2 times +
Predict the coordinates obtained in 2) and the coordinates obtained horizontally (2 times +2) and vertically (2 times +2) as effective elements in the next high-frequency subband space. Or the coordinates obtained by doubling the existing coordinates of the element horizontally and vertically, and the coordinates obtained by horizontally doubling and vertically (2 times +1), and horizontally (2 times +
1) and the coordinates obtained by vertically doubling, the coordinates obtained by horizontally (2 times +1) and vertically (2 times +1), and the horizontally (2 times -1) and vertically 2 The coordinates obtained by doubling,
Coordinates obtained by horizontally (twice -1) and vertically (twice +1), coordinates obtained by horizontally (twice +2) and vertically twice, and horizontally (twice +2) ) And vertically (2 times +1), and the coordinates obtained by horizontally doubling and vertically (2 times -1), and horizontally doubling and vertically (2 times +2)
And (2 times + 1) horizontally and vertically
A total of 12 element coordinates of the coordinates obtained by (2 times -1) and the coordinates obtained by horizontally (2 times +1) and vertically (2 times +2) are used as effective coordinates of the next high frequency subband space. It is possible to predict it as an element.

23A to 23E are explanatory views of the operation of the upsampler, where 2301 is a mask release point (a), 2302 is a mask release point (b), 2303 is a mask release point (c),
2304 is a mask release point (d), 2311 is a mask release point (e), 2312 is a mask release point (f), 2313 is a mask release point (g), 2314 is a mask release point (h), 2321
Is the mask release point (i), 2322 is the mask release point
(j), 2323 is a mask release point (k), 2324 is a mask release point (l), 2331 is a mask release point (m), 2332 is a mask release point (n), 2333 is a mask release point (o),
2334 is a mask release point (p).

In the first embodiment, the zero area prediction device 101
The prediction pattern of the mask removal point (a) shown in Fig. 23 (a)
2301 and the mask release point (b) 2302, the mask release point (c) 2303, and the mask release point (d) 2304. However, if the prediction threshold value is increased to increase the data compression rate, the image quality is degraded.

In order to solve this, in the second embodiment of the present invention, for an image including many high frequency components in the horizontal component, as shown in FIG. 23 (b), the mask release point (e) twenty three
11 and mask release point (f) 2312 and mask release point
A pattern obtained by adding (g) 2313 and mask release point (h) 2314 to the pattern of FIG. 23 (a) and expanding in the horizontal direction is used.
By doing so, the pattern is effective for an image including many high frequency components in the horizontal component.

Further, in the third embodiment, as shown in FIG. 23 (c), the mask release point (i) 2321 and the mask release point are applied to the image containing many high frequency components in the vertical component.
A pattern obtained by adding (j) 2322, the mask release point (k) 2323 and the mask release point (l) 2324 to the pattern of FIG. 23 (a) and extending in the vertical direction is used. By doing so, the pattern is effective for an image including many high frequency components in the vertical component.

Further, in the fourth embodiment, as shown in FIG. 23 (d), the mask release point (m) 2331 and the mask release point are applied to the image including many high frequency components in the oblique components.
A pattern obtained by adding (n) 2332, the mask release point (o) 2333 and the mask release point (p) 2334 to the pattern of FIG. 23 (a) and extending diagonally is used. By doing so, it is effective for an image including many high frequency components in the oblique component.

In the fifth embodiment, as shown in FIG. 23 (e), the mask release point (e) 2311 and the mask release point are set for an image containing various high frequency components.
(f) 2312, mask release point (g) 2313, mask release point (h) 2314, mask release point (i) 2321, mask release point (j) 2322, mask release point (k) 2323 and mask release point (l) ) 2324 is added to the pattern of FIG. 23 (a) and a pattern expanded in each direction is used. By doing so, it is effective for an image including many high frequency components in various components.

By the way, the mask plane space 506 of the HL subband space 1505 of the layer 2 and the mask plane space 509 of the HL subband space 1508 of the layer 3 include many frequency components in the horizontal component due to the inherent property of the subband space. There is a tendency. Also the layer 2 LH sub-band space
The mask plane space 507 of 1506 and the mask plane space 510 of LH subband space 1509 of layer 3 tend to include many frequency components in the vertical component. Further, the mask plane space 508 of the HH subband space 1507 of the layer 2 and the mask plane space 511 of the HH subband space 1510 of the layer 3 tend to include many frequency components in the diagonal components.

Therefore, in the sixth embodiment, the H of layer 2 is
Mask plane space 506 of L subband space 1505 and mask plane space of HL subband space 1508 of layer 3
The pattern of Fig. 23 (a) is used for 509, and the HL of layer 3 is used.
FIG. 23 shows the mask plane space 509 of the subband space 1508.
The pattern of (b) is used, and the mask plane space 507 and the layer 3 of the LH subband space 1506 of the layer 2 are used.
23 (c) is used for the mask plane space 510 of the LH subband space 1509, and the mask plane space 508 and the layer 3 of the HH subband space 1507 of the layer 2 are used.
23 (d) is used for the mask plane space 511 of the HH subband space 1510 of
It allows general images to be comprehensively compressed.

[0074]

As described above, according to the image data processing device and the image data processing method of the present invention, by predicting the zero value region of the low layer to the high layer in encoding and decoding, Data compression can be realized.

[Brief description of drawings]

FIG. 1 is a block diagram showing the structure of a wavelet image compression apparatus according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing the structure of a wavelet image expansion device according to the first embodiment of the present invention.

3 is a block diagram showing the structure of the zero region prediction device of FIG. 1. FIG.

FIG. 4 is a flowchart of a zero region prediction process according to the first embodiment of this invention.

FIG. 5 is an explanatory diagram of a mask plane releasing process and transfer data after mask processing according to the first embodiment of this invention.

FIG. 6 is a block diagram showing the structure of the zero area mask device of FIG. 1.

FIG. 7 is a flowchart of a zero area masking process according to the first embodiment of the present invention.

FIG. 8 is a block diagram showing the structure of the zero region interpolation device of FIG.

FIG. 9 is a flowchart of a zero area interpolation process according to the first embodiment of this invention.

FIG. 10 is a block diagram showing the structure of a conventional wavelet image compression device.

11 is a block diagram showing the structure of the wavelet transform device of FIG.

12 is a block diagram showing the structure of the two-dimensional filter bank of FIG.

13 is an explanatory diagram of filter characteristics in the two-dimensional filter bank in FIG.

FIG. 14 is an explanatory diagram of the operation of the down sampler.

FIG. 15 is an explanatory diagram of a wavelet space.

16 is a block diagram showing the structure of the code quantization apparatus of FIG.

17 is a block diagram showing the structure of the VLC of FIG.

18 is a block diagram showing the structure of the IVLC of FIG.

19 is a block diagram showing the structure of the code dequantization device of FIG.

20 is a block diagram showing the structure of the inverse wavelet transform device of FIG.

21 is a block diagram showing the structure of the inverse two-dimensional filter bank of FIG.

FIG. 22 is an explanatory diagram of the operation of the upsampler.

FIG. 23 is an explanatory diagram showing an extension mask release pattern in each direction.

[Explanation of symbols]

101 ... Zero region prediction device, 102 ... Zero region mask device, 12
1 ... Layer 2 region prediction data, 122 ... Layer 2 region prediction data, 123 ... Layer 2 quantization mask band data, 124 ... Layer 3 quantization mask band data,
201 ... Zero area interpolator, 221 ... Decoded layer 2
Quantization mask subband data, 222 ... Decoded layer 3 quantization mask subband data, 223 ... Decoded layer 3 zero area prediction data, 224 ... Decoded layer 2 zero area prediction data, 225 ... Decoded Layer 2 mask subband data, 226 ... Decoded layer 3 mask subband data, 301 ... Central processing unit for zero region prediction, 302 ... Subband data input port, 303 ... Prediction data output port, 304 ... Mask plane Memory, 305 ... Working memory for zero area prediction, 321 ... System bus for zero area prediction, S
401 to S411 ... Steps for zero region prediction, 501 ... Mask plane space corresponding to wavelet space, 50
2 ... Mask plane space of LL component subband space 1501 of layer 1, 503 ... Mask plane space of HL component subband space 1502 of layer 1, 504 ... Layer 1
LH component sub-band space 1503 mask plane space,
505 ... Mask plane space of HH component subband space 1504 of layer 1, 506 ... Mask plane space of HL component subband space 1505 of layer 2, 507 ... Mask plane space of LH component subband space 1506 of layer 2, 508 ... Layer 2 HH component subband space 1507
Mask plane space 509 ... Layer 3 HL component subband space 1508 mask plane space 510 ... Layer 3 LH component subband space 1509 mask plane space 511 ... Layer 3 HH component subband space 1
Mask plane space of 510, transfer source prediction source data in 521 ... 503, transfer target data in 522 ... 506, transfer target data in 523 ... 509, 551 ... 506
, The second transfer data at 552 ... 506, the third transfer data at 553 ... 506, the fourth transfer data at 554 ... 506, 555 ... 5
09 first transfer data, 556 ... 509 second transfer data, 557 ... 509 third transfer data, 558 ... 509 fourth transfer data, 559 ... 5
The fifth transfer data at 09, the sixth transfer data at 560 ... 509, the seventh transfer data at 561 ... 509, the eighth transfer data at 562 ... 509, 563
... 509 ninth transfer data, 564 ... 509 tenth transfer data, 565 ... 509 eleventh transfer data, 566 ... 509 twelfth transfer data,
601 ... Central processing unit for zero area mask, 602 ... Prediction / subband data input port, 603 ... Mask processing subband data output port, 604 ... Transfer buffer memory,
605 ... Working memory for zero area mask, 621 ... System bus for zero area mask, S701 to S711 ... Steps for zero area mask, 801 ... Central processing unit for zero area interpolation, 802 ... Mask -Subband data input port, 803 ... Interpolation processing Subband data output port, 804 ... Wavelet spatial memory, 805 ... Working memory for zero area interpolation, 821 ... System bus for zero area interpolation, S901 to S911 ... Steps for zero area interpolation, 1001 ... Input image, 1002 ... Wavelet transform device, 1003 ... Code quantizer, 1004 ... Variable length coding device, 1005 ... Variable length decoding device, 1006 ... Inverse code quantizing device, 1007 ... inverse wavelet transform device, 1008 ... decoded image, 1021 ... input image data, 1022 ... layer 1 subband data, 1023 ... layer 2 subband data, 1024 ... layer 3 subband Data, 1025 ... Layer 1 quantized subband data, 1026 ... Layer 2 quantized subband data, 1027 ... Layer 3 quantized subband data, 1028 ... Variable length compressed image data, 1029 ... Decoded Layer 1 quantized subband data Band data, 1030 ... Decoded layer 2 quantized subband data, 1031 ...
Decoded layer 3 quantized subband data, 1032
Decoded layer 1 subband data, 1033 ... Decoded layer 2 subband data, 1034 ... Decoded layer 3 subband data, 1035 ... Decoded image data, 1101 ... Two-dimensional filter bank, 1121 ... Layer 1 LL component subband data, 1122 ... Layer 1 LH component subband data, 1123 ... Layer 1 HL
Component subband data, 1124 ... Layer 1 HH component subband data, 1125 ... Layer 2 LH component subband data, 1126 ... Layer 2 HL component subband data, 1127 ... Layer 2 HH component subband data, 1128 ... LH component sub-band data of layer 3,
1129 ... HL component sub-band data of layer 3, 1130
... Layer 3 HH component sub-band data, 1131 ... Layer 3 LL component sub-band data, 1132 ... Layer 3 LL component sub-band data, 1201 ... Horizontal low-pass filter, 1202 ... Horizontal high-pass filter, 1203 ... Vertical low-pass filter, 1204… Vertical high-pass filter,
1205 ... horizontal downsampler, 1206 ... vertical downsampler, 1221 ... two-dimensional data, 1222 ... horizontal component low frequency subband data, 1223 ... horizontal component high frequency subband data, 1224 ... LL component subband data, 1225 ... L
H component sub-band data, 1226 ... HL component sub-band data, 1227 ... HH component sub-band data, 1301 ... Horizontal / vertical low pass filter frequency characteristics, 1302 ... Horizontal / vertical high pass filter frequency characteristics, 1401 ... Signal sample interval, 1402 ... Subsample interval, 1403 ... 1st
Sample 1404 ... second sample 1405 ... third sample 1406 ... fourth sample 1501 ... layer 1 LL component subband space 1502 ... layer 1 HL component subband space 1503 ... layer 1 LH component subband space of 1504, layer 1 HH component subband space, 1505, layer 2 HL component subband space, 1506, layer 2 LH component subband space,
1507 ... Layer 2 HH component subband space, 1508 ...
Layer 3 HL component subband space, 1509 ... Layer 3 LH component subband space, 1510 ... Layer 3 HH component subband space, 1601 ... Integer divider, 1602
... near zero clipping device, 1621 ... layer 1 LL component quantized subband data, 1622 ... clip unprocessed layer 1 LH component quantized subband data, 1623 ...
HL component quantized sub-band data of layer 1 without clip processing, 1624 ... HH of layer 1 without clip processing
Component quantized sub-band data, 1625 ... LH component quantized sub-band data of layer 2 that is not clipped
26 ... HL component quantized subband data of unprocessed layer 2, 1627 ... HH component quantized subband data of unprocessed layer 2, 1628 ... LH component quantized subband data of unprocessed layer 3 ,
1629 ... HL component quantized sub-band data of clip-unprocessed layer 3, 1630 ... clip-unprocessed layer 3
HH component quantized sub-band data, 1631 ... Layer 1
LH component quantized subband data of 1632, layer 1 HL component quantized subband data, 1633 ... layer 1 of HH component quantized subband data, 1634 ... layer 2 of LH component quantized subband data, 1635. Layer 2 HL component quantized sub-band data, 1636 ...
Layer 2 HH component quantized subband data, 1637
... Layer 3 LH component quantized sub-band data, 16
38 ... HL component quantized subband data of layer 3,
1639 ... HH component quantized sub-band data of layer 3,
1701 ... P-S converter, 1702 ... lossless encoder,
1721 ... Compressed continuous data, 1801 ... Reversible decoding device,
1802 ... SP converter, 1821 ... Expanded continuous data, 18
22 ... Decoded layer 1 LL component quantized subband data, 1823 ... Decoded layer 1 LH component quantized subband data, 1824 ... Decoded layer 1 HL component quantized subband data, 1825 ... Decoded layer 1 HH component quantized subband data, 1826 ...
Decoded layer 2 LH component quantized subband data, 1827 ... Decoded layer 2 HL component quantized subband data, 1828 ... Decoded layer 2 HH
Component quantized subband data, 1829 ... Decoded layer 3 LH component quantized subband data, 1830 ... Decoded layer 3 HL component quantized subband data, 1831 ... Decoded layer 3 HH component quanta 1901 ... Multiplier, 1921 ... Decoded layer 1 LL component subband data, 1922 ... Decoded layer 1 LH component subband data, 1923
... Decoded layer 1 HL component subband data,
1924 ... Decoded layer 1 HH component subband data, 1925 ... Decoded layer 2 LH component subband data, 1926 ... Decoded layer 2 HL component subband data, 1927 ... Decoded layer 2 H
H component subband data, 1928 ... Decoded layer 3 LH child subband data, 1929 ... Decoded layer 3 HL component subband data, 1930 ... Decoded layer 3 HH component subband data, 2001 …
Inverse two-dimensional filter, 2021 ... Decoded layer 2 L
L component sub-band data, 2022 ... Decoded layer 3 LL component sub-band data, 2101 ... Band synthesis filter based on horizontal low-pass filter, 2102 ... Band synthesis filter based on horizontal high-pass filter, 2103 ... Vertical low-pass filter Band synthesis filter based on 21
04 ... Band synthesis filter based on vertical high-pass filter, 2105 ... Horizontal upsampler, 2121 ... LL component subband data, 2122 ... LH component subband data,
2123 ... HL component subband data, 2124 ... HH component subband data, 2125 ... Horizontal component low frequency subband data, 2126 ... Horizontal component high frequency subband data,
2127 ... Two-dimensional data, 2201 ... Signal sample interval, 22
02… upsampling interval, 2203… first sample,
2204 ... 2nd sample, 2205 ... 1st 0 insertion, 2206
… Second 0 insertion, 2301… Mask release point a, 230
2… Mask release point b, 2303… Mask release point
c, 2304 ... Mask release point d, 2311 ... Mask release point e, 2312 ... Mask release point f, 2313 ... Mask release point g, 2314 ... Mask release point h,
2321 ... Mask release point i, 2322 ... Mask release point j, 2323 ... Mask release point k, 2324 ... Mask release point l, 2331 ... Mask release point m, 2332
… Mask release point n, 2333… Mask release point
o, 2334 ... Mask release point p.

Claims (8)

[Claims] 1. A null region predictor having means for predicting a significant space of a high frequency subband from a low frequency subband of a wavelet space, and a means for masking data other than the significant space predicted by the null region predictor. An image data processing device, comprising: a zero area mask device having the same. 2. A zero region predicting device having means for predicting a significant space of a high frequency subband from a low frequency subband of a wavelet space, and only a significant space predicted by the zero region predicting device for wavelet-transformed data. And a zero area interpolating device having means for allocating to the image data processing device. 3. When an element in the low-frequency subband space is larger than an arbitrary value, the coordinates at which the element exists are doubled horizontally and vertically and doubled horizontally. And the coordinates obtained by vertically (twice + 1) and horizontally
The coordinates obtained by (2 times +1) and vertically doubling, and the coordinates obtained by horizontally (2 times +1) and vertically (2 times +1), a total of 4 element coordinates, are used as the next high frequency subband. An image data processing method characterized by predicting as an effective element of space. 4. When an element in the low-frequency subband space is larger than an arbitrary value, the coordinate where the element exists is set to 2 horizontally.
Doubled and vertically doubled, the coordinates doubled horizontally and vertically (twice +1), and the horizontal (2
X +1) and the coordinates obtained by doubling vertically and horizontally
Coordinates obtained by (2 times +1) and vertically (2 times +1), and coordinates obtained by horizontally (2 times -1) and vertically doubling, and horizontally (2 times -1) And the coordinates obtained by vertically (2 times + 1), the coordinates obtained by horizontally (2 times + 2) and vertically twice, and the horizontally (2 times + 2) and vertically (2 times + 1) )
An image data processing method, comprising: predicting a total of eight element coordinates of the coordinates obtained as described above as effective elements of the next high frequency subband space. 5. When an element in the low-frequency subband space is larger than an arbitrary value, the coordinates of the element are set to 2 horizontally.
Doubled and vertically doubled, the coordinates doubled horizontally and vertically (twice +1), and the horizontal (2
X +1) and the coordinates obtained by doubling vertically and horizontally
(2 times +1) and vertically (2 times +1), the coordinates obtained by doubling horizontally and vertically (2 times -1), and the coordinates obtained by doubling horizontally and vertically ( Coordinates obtained by multiplying (2 times +2), coordinates obtained by horizontally (2 times +1) and vertically (2 times -1), and coordinates obtained by horizontally (2 times +1) and vertically (2 times +
An image data processing method, characterized in that a total of 8 element coordinates of the coordinates obtained in 2) are predicted as effective elements of the next high frequency subband space. 6. When an element in the low-frequency subband space is larger than an arbitrary value, the coordinates of the element are set to 2 horizontally.
Doubled and vertically doubled, the coordinates doubled horizontally and vertically (twice +1), and the horizontal (2
X +1) and the coordinates obtained by doubling vertically and horizontally
(2 times +1) and vertically obtained (2 times +1), and horizontally obtained (2 times -1) and vertically obtained (2 times -1). Times +2) and vertically (2 times -1), and horizontally (2 times -1) and vertically (2 times -1)
Predicted as a valid element of the next high-frequency subband space, a total of 8 element coordinates of the coordinates obtained by multiplying (+2) times and the coordinates obtained by horizontally (2 times +2) and vertically (2 times +2). An image data processing method characterized by: 7. When an element in the low-frequency subband space is larger than an arbitrary value, the coordinates of the element are set to 2 horizontally.
Doubled and vertically doubled, the coordinates doubled horizontally and vertically (twice +1), and the horizontal (2
X +1) and the coordinates obtained by doubling vertically and horizontally
Coordinates obtained by (2 times +1) and vertically (2 times +1), and coordinates obtained by horizontally (2 times -1) and vertically doubling, and horizontally (2 times -1) And the coordinates obtained vertically (2 times + 1), the coordinates obtained horizontally (2 times + 2) and vertically twice, and the coordinates obtained horizontally (2 times + 2) and vertically (2 times +
The coordinate obtained in 1) and the coordinate obtained by doubling horizontally and vertically (2 times-1) and the coordinate obtained by doubling horizontally and vertically (2
Coordinates obtained by multiplying (+2) times, coordinates obtained by horizontally (2 times +1) and vertically (2 times -1), coordinates obtained by horizontally (2 times +1) and vertically (2 times +2) Total of the coordinates obtained by 1
An image data processing method characterized by predicting a two-element coordinate as an effective element in the next high-frequency subband space. 8. The image data according to claim 3, wherein a prediction pattern of effective elements is selected for the essential horizontal frequency characteristic and vertical frequency characteristic of each subband space. Processing method.