JPH09107548A - Image compression device and image compression method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent deterioration in image quality even for an image in a rapid motion and to attain compression with a high compression rate by making quantization coefficient data for a high frequency band corresponding to an image area where a motion is rapid and the motion scalar is high. SOLUTION: A wavelet transformation section 3 converts image data received from a frame memory 1 via a selector 2 into coefficient data for each frequency band. A quantization section 4 quantizes the coefficient data for each sub band. A inverse quantization section 5 and a wavelet transformation section 6 decode image data and a motion compensation section 9 conducts forward inter-frame prediction coding from difference data with a current image. A quantization coefficient nullifying section 10 discriminates whether or not the data are high frequency band data based on the quantization coefficient data and when motion scalar in the longitudinal and lateral directions obtained by the motion compensation section 9exceeds a prescribed amount stored in a quantization coefficient nullifying motion quantity table 11, the quantization coefficient data are nullified. Thus, the compression rate is increased without giving effect onto the image quality.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression apparatus and an image compression method for reducing an amount of transmitted information by compressing an image.

[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. But DCT
As an essential problem inherent in the compression method using, there is a problem that when the compression rate is increased, block distortion, mosquito noise, etc. are visually recognized and the compression rate is 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, has been receiving attention.
When the wavelet transform is used, since there is no concept of block, the block distortion generated in DCT is eliminated,
Visually significant image quality improvement can be expected.

Description will be made by comparing a compressed image by DCT and a compressed image by a compression technique using wavelet transform. First, a DCT-compressed image 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 components are preserved, but the distortion of the visually-prominent high-frequency components occurs. On the other hand, the compressed image obtained by the compression technique using the wavelet transform becomes an analog image, and the high frequency component is naturally lost by increasing the compression rate. In other words, the compressed image signal is gradually cut from the high frequency part of the signal band. As a result, the resolution is reduced as a whole. Therefore, if the compression ratio is the same,
It cuts out more high-frequency components that are visually distorted.
The wavelet transform causes less visual deterioration than the compressed image by DCT.

Conventionally, as an image compression apparatus, the one shown in FIG. 7 is known. In FIG. 7, 101 is 1
A frame memory that stores the input image data of the frame,
Reference numeral 102 denotes a selector capable of connecting two terminals out of three terminals (thus, the number of connections is three), and 103 reduces the image data output from the frame memory 101 to a half by a low-pass filter and a high-pass filter. A wavelet transform unit for obtaining coefficient data by thinning each band-divided image data by half to reduce the amount of image data (performing wavelet transform), and a wavelet transform unit 104. A quantizing unit that quantizes coefficient data output from the unit 103 for each subband (each frequency band) to obtain quantized coefficient data, and 105 represents quantized coefficient data output from the quantizing unit 104. Is restored to wavelet transform coefficient data, and 106 is the wavelet transform coefficient data to image data. Inverse wavelet transform unit which, 10
7 is an adder for adding two inputs, and 108 is a quantizer 104
A variable length coding unit 109 for reducing the information amount of the entire image data output from the input 109, the restored image data and the current image data are input, and the restored image data is divided into a plurality of blocks including a plurality of image data. The vertical and horizontal distances that minimize the difference between the in-block constituent pixels of the divided and restored image and the in-block constituent pixels of the current image data are extracted for each divided block, and the inter-prediction between the forward frames is predicted by the difference data. This is a motion compensation unit that performs encoding.

The operation of the conventional image compression apparatus configured as described above will be described below. First, the compression processing of image data will be described. The image data is input to the frame memory 101. In the intraframe coding process, the selector 102 connects the output side of the frame memory 101 and the input side of the wavelet transform unit 103, and the image data from the frame memory 101 is input to the wavelet transform unit 103.

FIG. 8 is a block diagram showing arithmetic processing of wavelet transform. As shown in FIG. 8, the input image data to the wavelet transform unit 103 includes horizontal low-pass filter (horizontal LPF) and high-pass filter (horizontal H).
PF) and divides the band of the image data into two, and the half subsamplers 110a and 110b decimate the data amount to half each. Next, half subsampler 1
The image data output from 10a includes a vertical low-pass filter (vertical LPF) and a high-pass filter (vertical HP).
F), the bandwidth of the image data is divided into two, and the amount of data is thinned out to half by the half subsamplers 111a and 111b. Also, half subsampler 11
The image data output from 0b is a low-pass filter (vertical LPF) in the vertical direction and a high-pass filter (vertical HP).
F), the bandwidth of the image data is divided into two, and the amount of data is thinned to half by the half subsamplers 111c and 111d. At this time, the output image data from the half subsamplers 111b, 111c, 111d are no longer processed, and the frequency bands W1LH, W1HL,
Although it becomes the coefficient data of W1HH, the same processing as above is repeated for the image data output from the half subsampler 11a. That is, the component of the image data (horizontal L
The same processing as described above is repeated for the image data component band-limited by the PF and band-limited by the vertical LPF. As a result of the repetition of such processing, the coefficient data generated as a result is stored as coefficient data in which the amount of frequency division in the horizontal and vertical directions is reduced by half along the low frequency region. Will be things.

FIG. 9 is a data state diagram showing wavelet transform coefficient data for each of a plurality of frequency bands. FIG.
Indicates a state in which the wavelet transform is performed up to the third time for convenience. In this way, the wavelet-transformed coefficient data is distributed in the horizontal and vertical directions to form a hierarchical structure. The above was about compression processing,
In the decompression process, the high frequency component is added to the low frequency component by decoding the image of the four regions of W3 in FIG. 9 into the image of W1 (by performing the process in the opposite direction to that of FIG. 8). By superimposing, it is possible to realize the stepwise improvement of resolution.

The quantizing unit 104 quantizes the coefficient data for each frequency band converted by the wavelet transforming unit 103, and the variable length coding unit 108 outputs the quantum output from the quantizing unit 104. The information amount of the entire data is reduced by allocating more information to the data having a higher occurrence probability among the conversion coefficient data. In this way, the intraframe coding process (intraframe image compression process)
Is performed.

In the inter-frame coding process, the data quantized by the quantizing unit 104 is decoded into the coefficient data which is wavelet transformed by the inverse quantizing unit 105, and the wavelet inverse transforming unit 106 restores the image data, The restored previous image data and the current image data input to the frame memory 101 are input to the motion compensation unit 109 by switching the selector 102, and the motion component is extracted. Specifically, the previous image data is divided into blocks composed of a plurality of pixels, and horizontal and vertical vector data that minimizes the difference data between the current image data and the previous image data is obtained for each block. Extract. The motion compensation unit 109 collects the above difference data for one screen (for one frame) and outputs it to the wavelet transform unit 103 via the selector 102. In the wavelet transform unit 103 and thereafter, the same coding process as the intraframe coding process is performed. I do.

By the way, the compression rate of image data depends on the compression rate in inter-frame coding rather than in frame,
In the inter-frame coding, motion compensation is performed so as to minimize the difference data, but in the conventional inter-frame coding process, it is difficult to completely compensate the motion for an image region having a lot of motion. Therefore, the difference data becomes large in the image data in which there are many areas with a large amount of motion due to insufficient motion compensation, and the compression rate is reduced accordingly. The fact that the difference data of the image data with many areas with large movements becomes large means that the deterioration of resolution due to visual characteristics does not matter in the areas with large movements on the screen, but it does not improve image quality due to visual characteristics. This means that a useless amount of data is given to a portion that does not contribute and has a large amount of movement.

[0011]

In this image compression apparatus, it is required that the compression rate does not decrease even for image data in which there are many areas in which the motion is vigorous.

According to the present invention, an image compressing device capable of preventing deterioration of image quality and performing high compression even for image data having a lot of dynamic regions, and image quality of image data having a lot of dynamic regions. It is an object of the present invention to provide an image compression method capable of preventing deterioration of the image quality and performing high compression.

[0013]

In order to solve this problem, the present invention uses a frame memory for storing input image data of one frame and image data output from the frame memory by a low-pass filter and a high-pass filter. A wavelet transform unit that obtains coefficient data by performing a predetermined number of steps to reduce the amount of image data by thinning out each image data that is divided into halves into two.
A quantizing unit that quantizes coefficient data output from the wavelet transform unit for each sub-band to obtain quantized coefficient data, and a quantized coefficient 0 that stores a motion scalar amount that zeroizes the quantized coefficient. Table, the inverse quantization unit that restores the quantized coefficient data output from the quantization unit to the coefficient data of the wavelet transform, and the wavelet inverse that restores the coefficient data restored by the inverse quantization unit to the image data. A conversion unit, which inputs the restored image data and the current image data, divides the restored image data into a plurality of blocks composed of a plurality of images, and configures a pixel in a block of the restored image and an in-block configuration of the current image data. A motion in which the vertical and horizontal distances that minimize the difference from the pixel are extracted for each of the divided blocks, and the forward interframe prediction coding is performed based on the difference data. The compensation unit and the motion scalar amount in the quantized coefficient zeroized motion amount table are compared with the motion scalar amount that is the distance data in the vertical and horizontal directions in the motion compensation unit, and the motion that is the distance data in the vertical and horizontal directions is compared. When the scalar amount exceeds a predetermined amount and the corresponding quantized coefficient data is high frequency band data, a quantized coefficient zeroizing unit that sets the quantized coefficient data to 0 for the corresponding image area,
It is configured to have a variable length coding unit that allocates more information to data having a higher occurrence probability among the quantized coefficient data output from the quantized coefficient zeroization unit.

As a result, it is possible to obtain an image compression apparatus capable of preventing deterioration of image quality and performing high compression even for image data in which a lot of areas are moving rapidly.

[0015]

BEST MODE FOR CARRYING OUT THE INVENTION According to the first aspect of the present invention, a frame memory for storing one frame of input image data and image data output from the frame memory are divided into two parts by a low-pass filter and a high-pass filter. Output from the wavelet transform unit that obtains coefficient data by performing a predetermined number of steps to reduce the image data amount by thinning out each image data band-divided into 1 A quantization unit that quantizes coefficient data for each sub-band to obtain quantized coefficient data; a quantized coefficient zero motion amount table that stores a motion scalar amount that zeroizes the quantized coefficient; The inverse quantization unit that restores the quantized coefficient data output from the quantization unit to the coefficient data of the wavelet transform, and the coefficient data restored by the inverse quantization unit Enter the wavelet inverse transformation unit to restore, the restored image data and current image data,
The restored image data is divided into a plurality of blocks composed of a plurality of images, and the distances in the vertical and horizontal directions that minimize the difference between the in-block constituent pixels of the restored image and the in-block constituent pixels of the current image data are A motion compensation unit that extracts each divided block and performs forward interframe predictive coding based on the difference data, a motion scalar amount in the quantized coefficient 0 motion amount table, and vertical and horizontal distance data in the motion compensation unit. When a certain amount of motion scalar is compared and the amount of motion scalar that is distance data in the vertical and horizontal directions exceeds a predetermined amount and the corresponding quantization coefficient data is high frequency band data, the corresponding image area is quantized. The quantized coefficient zeroization unit that sets the coefficient data to 0 and the quantized coefficient data output from the quantized coefficient zeroization unit that has a higher probability of occurrence. And a variable length coding unit for allocating more information to a corresponding block when the amount of motion scalar between frames exceeds a predetermined amount and the corresponding quantization coefficient data is high frequency band data. With respect to the image area of, the quantized coefficient data is set to 0.

According to a second aspect of the present invention, a frame memory for storing one frame of input image data, and image data output from the frame memory are band-divided into halves by a low-pass filter and a high-pass filter, respectively. Of the image data is thinned out by half to reduce the amount of image data, and a predetermined number of steps are performed to obtain coefficient data. A quantizer that quantizes each band to obtain quantized coefficient data; a subband pair quantized coefficient zero motion amount table that stores a motion scalar amount that zeroizes the quantized coefficient for each subband; The inverse quantization unit that restores the quantized coefficient data output from the quantization unit to the coefficient data of the wavelet transform, and the coefficient data restored by the inverse quantization unit The wavelet inverse transform unit for restoring the image data, the restored image data and the current image data are input, the restored image data is divided into a plurality of blocks composed of a plurality of images, and the constituent pixels in the block of the restored image and the current image data are divided. A motion compensating unit that performs forward interframe predictive coding by extracting the vertical and horizontal distances that minimize the difference between the image data and the constituent pixels in the block, and the subband pair quantum The motion scalar amount in the motion amount table and the motion scalar amount that is the distance data in the vertical direction and the horizontal direction in the motion compensation unit are compared for each subband, and the motion scalar that is the distance data in the vertical direction and the horizontal direction is compared. When the amount exceeds the predetermined amount for each sub-band, the quantization coefficient zeroization unit that sets the quantization coefficient data to 0 for the corresponding image area, and the amount The quantized coefficient data output from the quantized coefficient zeroization section has a variable length coding section that allocates more information to data having a higher occurrence probability, and the amount of motion scalar between frames is sub When the amount exceeds the predetermined amount for each band, the quantization coefficient data is set to 0 for the image data of the image area of the corresponding block.

According to the third aspect of the invention, the image data amount is reduced by halving each image data obtained by band-dividing the image data into halves by the low-pass filter and the high-pass filter. A wavelet transformation step of performing a predetermined number of steps to obtain coefficient data, and a quantization step of obtaining quantized coefficient data by performing quantization for each subband on the coefficient data output from the wavelet transformation unit, Inverse quantization step to restore the quantized coefficient data to coefficient data of the wavelet transform, wavelet inverse transform step to restore the coefficient data restored in the inverse quantization step to image data, restored image data and current image data , The restored image data is divided into multiple blocks consisting of multiple images, and the blocks of the restored image Difference constituent pixel and a block arrangement of pixels of the current image data is extracted for each block obtained by dividing a distance in the vertical direction and the lateral direction becomes minimum and a motion compensation step of performing inter front frame predictive coding by the difference data,
The held motion scalar amount is compared with the motion scalar amount, which is the distance data in the vertical and horizontal directions in the motion compensator, and when the motion scalar amount, which is the distance data in the vertical and horizontal directions, exceeds a predetermined amount. The corresponding image area has a quantized coefficient zeroization step for setting the quantized coefficient data to 0, and a variable length coding step for allocating more information to the quantized coefficient data having a higher occurrence probability. When the amount of motion scalar between frames exceeds a predetermined amount and the corresponding quantized coefficient data is high frequency band data, the quantized coefficient data is set to 0 for the image area of the corresponding block. Have.

According to a fourth aspect of the present invention, in the third aspect of the present invention, the held motion scalar quantity is compared with the motion scalar quantity which is distance data in the vertical and horizontal directions in the motion compensator. When the amount of motion scalar, which is the distance data in the vertical and horizontal directions, exceeds a predetermined amount, the corresponding image area is held instead of the quantization coefficient zeroization step in which the quantization coefficient data is set to zero. The motion scalar amount, which is the distance data in the vertical and horizontal directions in the motion compensation unit, is compared for each subband, and the motion scalar amount that is the distance data in the vertical and horizontal directions is a predetermined amount for each subband. When the value exceeds the value, a quantized coefficient zero step for setting the quantized coefficient data to 0 is provided for the corresponding image area, and the amount of motion scalar between frames is reduced. When exceeding the predetermined amount of each band has an advantage in that the zero quantized coefficient data for the image data in the image area of the corresponding block.

(Embodiment 1) An embodiment of the present invention will be described below with reference to FIGS. 1 to 6.

FIG. 1 is a block diagram showing an image compression apparatus according to an embodiment of the present invention. In FIG. 1, the frame memory 1 stores one frame of input image data.
The selector 2 enables connection between two terminals out of three terminals (thus, there are three kinds of connections). The wavelet transform unit 3 reduces the amount of image data by thinning out the image data output from the frame memory 1 into halves by dividing the image data into halves by a low pass filter and a high pass filter. A predetermined number of steps of processing (wavelet transformation) are performed to obtain coefficient data. The quantization unit 4 quantizes the coefficient data output from the wavelet transform unit 3 for each subband (each frequency band) to obtain quantized coefficient data. The inverse quantization unit 5 restores the quantized coefficient data output from the quantization unit 4 into the wavelet transform coefficient data (inverse quantization step). The wavelet inverse transform unit 6 restores the wavelet transform coefficient data to image data (inverse transform step). The adder 7 adds two inputs. Variable length coding unit 8
Aims to reduce the information amount of the entire image data by allocating more information to the data having a higher occurrence probability among the quantized coefficient data output from the quantized coefficient zeroization unit 10 described later (variable length coding step). ). The motion compensation unit 9 inputs the restored image data and the current image data, divides the restored image data into a plurality of blocks composed of a plurality of image data, and configures the constituent pixels in the block of the restored image and the block of the current image data. The vertical and horizontal distances that minimize the difference from the internal constituent pixels are extracted for each divided block, and forward interframe predictive coding is performed using the difference data (motion compensation step). The quantized coefficient 0 conversion unit 10 compares the motion scalar amount in the quantized coefficient 0 motion amount table 11 described later with the motion scalar amount output from the motion compensation unit 9, and the image region in which the motion scalar amount is larger than a predetermined amount. For, the quantized coefficient data is set to 0. The quantized coefficient zeroized motion amount table 11 stores a motion scalar amount that zeroizes the quantized coefficient.

The operation of the image compression apparatus configured as described above will be described below with reference to FIGS. 2 and 3. FIG. 2 is a flow chart for explaining the operation of the image compression apparatus of FIG. 1, and FIG. 3 is a data state diagram showing the motion scalar amount stored in the quantization coefficient zeroized motion amount table 11. First, the wavelet transform unit 3 performs the wavelet transform on the image data obtained from the frame memory 1 via the selector 2, so that the image data is converted into the coefficient for each frequency band as described in the section of the related art. Convert to data (S1, conversion step). Next, the quantizer 4 quantizes the coefficient data for each frequency band converted by the wavelet transformer 3 to obtain quantized coefficient data (S2, quantization step). Next, the quantized coefficient zeroization unit 10 determines whether or not the quantized coefficient data output from the quantization unit 4 is high frequency band data (S).
3, quantization coefficient 0 conversion step). In the case of the quantized coefficient data in the high frequency band, it is next determined whether or not the corresponding motion scalar quantity exceeds the motion scalar quantity (predetermined quantity) stored in the quantized coefficient 0-ized motion quantity table 11 ( (S4, quantized coefficient 0 conversion step), and when it is determined that the value is exceeded, the quantized coefficient data is set to 0 (S5, quantized coefficient 0 conversion step). When the quantized coefficient data to be determined is not the data in the high frequency band or when the motion amount corresponding to the quantized coefficient data does not exceed the predetermined amount, the variable length coding unit 8 sets the quantized coefficient 0 unit. The non-zeroed quantized coefficient data output from is subjected to variable length coding (variable length coding step). That is, the quantized coefficient data is set to 0 only when the quantized coefficient data is high frequency band data and the motion amount corresponding to the quantized coefficient data exceeds a predetermined amount.

FIG. 3 shows a quantized coefficient 0 motion amount table 1
1 shows the amount of motion scalar and the quantized coefficient stored in 1. FIG.
Then, the motion scalar amount is “16”, and this value is a predetermined amount that sets the quantized coefficient data to 0. When the motion scalar amount assigned to each quantized coefficient data exceeds “16”, Sets the quantization coefficient data to 0 if it is high frequency band data.

As described above, according to the present embodiment, since the quantization coefficient data in the high frequency band corresponding to the image area in which the motion is intense and the amount of scalar is large is 0, the resolution that does not affect the image quality is lowered. Therefore, it is possible to realize an image compression device capable of increasing the compression rate without degrading the image quality.

(Second Embodiment) FIG. 4 shows an image compression apparatus according to a second embodiment of the present invention. In FIG.
The frame memory 1, the selector 2, the wavelet transform unit 3, the quantization unit 4, the inverse quantization unit 5, the wavelet inverse transform unit 6, the adder 7, the variable length coding unit 8, and the motion compensation unit 9 are the same as those in FIG. Since it is a thing, the description is omitted. Quantization coefficient 0
The quantization unit 12 includes a motion scalar amount and a motion compensation unit 9 stored in a subband-to-quantization coefficient zeroized motion amount table 13, which will be described later.
It is compared with the amount of motion scalar output from each of the sub-bands to determine whether or not the quantized coefficient data is set to 0 for each sub-band. The subband-to-quantization coefficient zeroized motion amount table 13 stores a motion scalar amount as a predetermined amount for zeroizing the quantized coefficient data for each subband.

The operation of the image compression apparatus configured as described above will be described below with reference to FIGS. 5 and 6. FIG. 5 is a flow chart for explaining the operation of the image compression apparatus of FIG. 4, and FIG. 6 is a data state diagram showing the motion scalar amount stored in the subband-to-quantization coefficient zeroized motion amount table 13. First, the wavelet transform unit 3 performs the wavelet transform on the image data obtained from the frame memory 1 via the selector 2, so that the image data is converted into the coefficient for each frequency band as described in the section of the related art. Convert to data (S11, conversion step). Next, the quantizer 4 quantizes the coefficient data for each frequency band converted by the wavelet transformer 3 to obtain quantized coefficient data (S12, quantization step).

Next, the quantized coefficient zeroization section 10 has a scalar quantity (motion scalar quantity) of the motion vector obtained from the motion compensation section 9 for all the quantized coefficient data output from the quantization section 4. ) And the amount of motion scalar (predetermined amount) stored in the subband pair quantized coefficient zeroized motion amount table 13 are compared for each subband (frequency band) to determine whether the quantized coefficient data is zeroized data. Whether or not, that is, whether or not the amount of motion scalar assigned to the quantized coefficient data exceeds the predetermined amount determined for each subband is determined for each subband (S13, quantized coefficient 0 conversion step). When it is determined that the amount of motion scalar assigned to the quantized coefficient data exceeds the predetermined amount for each subband, the quantized coefficient data corresponding to the determined subband is set to 0.
(S14, quantization coefficient 0 conversion step). When it is determined that the amount of motion scalar assigned to the quantized coefficient data does not exceed the predetermined amount, the variable-length coding unit 8 outputs the quantized coefficient data that has not been quantized and is output from the quantized coefficient 0 quantized unit 12. Variable length coding (variable length coding step).

FIG. 6 shows the motion scalar quantity and the quantized coefficient stored in the subband pair quantized coefficient zeroized motion amount table 13. In FIG. 6, the amount of motion scalar is determined for each subband. When the amount of motion scalar is exceeded, the quantized coefficient data corresponding to that subband is set to zero. For example, the motion scalar amount of the subband W2 is “24”, and the motion scalar amount of the image data corresponding to the subband W2 is “2”.
When it exceeds 4 ", the quantized coefficient data corresponding to the subband W2 becomes 0. In the case of the quantized coefficient data corresponding to the subband W2, even if the motion scalar amount exceeds" 24 ", If it is within “32”, it does not become 0. In this way, the reference motion scalar quantity (predetermined quantity) is determined stepwise according to the height of the frequency of the frequency band (for example, the center frequency), and the quantization coefficient data is set to 0. Whether or not to do so is determined stepwise according to each frequency band.

As described above, according to the present embodiment, the coefficient data of each frequency band of the image area in which the motion is intense and the amount of scalar is large can be gradually set to 0. It is possible to realize the image compression device that can set the quantization coefficient data to 0 more accurately than in the case of uniformly performing the processing, and can further increase the compression rate without degrading the image quality.

[0029]

As described above, according to the present invention, since the quantization coefficient data in the high frequency band corresponding to the image region in which the motion is intense and the amount of scalar is large can be set to 0, the resolution that does not affect the image quality can be obtained. However, the compression rate can be increased without deteriorating the image quality, and the coefficient data of each frequency band of the image region in which the motion is large and the amount of scalar is large can be set to 0 stepwise. As compared with the case where the processing is uniformly performed on each image area, the quantization coefficient data can be more accurately set to 0, and the advantageous effect that the compression rate can be further increased without degrading the image quality is obtained. can get.

[Brief description of the drawings]

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

FIG. 2 is a flowchart for explaining the operation of the image compression apparatus in FIG.

FIG. 3 is a data state diagram showing a motion scalar amount stored in a quantization coefficient 0 motion amount table.

FIG. 4 is a block diagram showing an image compression apparatus according to a second embodiment of the present invention.

5 is a flowchart for explaining the operation of the image compression apparatus in FIG.

FIG. 6 is a data state diagram showing a motion scalar amount stored in a subband-to-quantization coefficient-zeroized motion amount table.

FIG. 7 is a block diagram showing a conventional image compression device.

FIG. 8 is a block diagram showing a calculation process of wavelet transform.

FIG. 9 is a data state diagram showing wavelet transform coefficient data for each of a plurality of frequency bands.

[Explanation of symbols]

 1 Frame Memory 2 Selector 3 Wavelet Transform Unit 4 Quantization Unit 5 Inverse Quantization Unit 6 Wavelet Inverse Transform Unit 7 Adder 8 Variable Length Encoding Unit 9 Motion Compensation Unit 10, 12 Quantization Coefficient 0 Quantization Coefficient 0 Quantization Coefficient 0 Motion amount table 13 subband pair quantization coefficient 0 bit motion amount table

Claims (4)

[Claims] 1. A frame memory for storing one frame of input image data, and image data output from the frame memory by a low pass filter and a high pass filter.
1/2 of each image data divided into half
The wavelet transform unit that obtains coefficient data by performing a predetermined number of steps for thinning out the image data amount and performing quantization for each subband on the coefficient data output from the wavelet transform unit. And a quantized coefficient motion amount table that stores a motion scalar amount that zeroizes the quantized coefficient, and a quantized coefficient data output from the quantized unit is wavelet transformed. The inverse quantizer for restoring the coefficient data, the wavelet inverse transform unit for restoring the coefficient data restored by the inverse quantizer into image data, the restored image data and the current image data are input, and The restored image data is divided into a plurality of blocks composed of a plurality of images, and the in-block constituent pixels of the restored image and the in-block constituent pixels of the current image data are divided. In the vertical direction and the horizontal direction that minimize the difference between the divided blocks, and a motion compensation unit that performs forward interframe predictive coding based on the difference data; The motion scalar amount and the motion scalar amount, which is the distance data in the vertical and horizontal directions in the motion compensator, are compared, and the motion scalar amount that is the distance data in the vertical and horizontal directions exceeds a predetermined amount and the corresponding quantum When the quantized coefficient data is the data in the high frequency band, the quantized coefficient zeroizing unit that sets the quantized coefficient data to 0 for the corresponding image area, and the quantized coefficient data output from the quantized coefficient zeroizing unit An image compression apparatus configured to include a variable length coding unit that allocates more information to data having a higher occurrence probability. 2. A frame memory for storing one frame of input image data, and image data output from the frame memory by a low pass filter and a high pass filter.
1/2 of each image data divided into half
The wavelet transform unit that obtains coefficient data by performing a predetermined number of steps for thinning out the image data amount and performing quantization for each subband on the coefficient data output from the wavelet transform unit. Output from the quantizer, and a quantizer for obtaining quantized coefficient data, a subband-to-quantized coefficient zero motion amount table for storing a motion scalar quantity for zeroizing the quantized coefficient for each subband, An inverse quantization unit that restores quantized coefficient data to coefficient data of wavelet transform, a wavelet inverse transform unit that restores the coefficient data restored by the inverse quantization unit to image data, the restored image data and the current image Data is input, the restored image data is divided into a plurality of blocks composed of a plurality of images, and the constituent pixels in a block of the restored image and the current image are divided. A distance in the vertical direction and the horizontal direction that minimizes the difference from the in-block constituent pixels of each data block for each of the divided blocks, and performs a predictive interframe prediction coding based on the difference data; The amount of motion scalar in the band-to-quantization coefficient-zero motion amount table and the amount of motion scalar that is distance data in the vertical and horizontal directions in the motion compensator are compared for each sub-band, and the vertical and horizontal distances are compared. When the amount of motion scalar, which is data, exceeds the predetermined amount for each sub-band, the quantized coefficient zeroizing unit for setting the quantized coefficient data to 0 for the corresponding image area is output from the quantized coefficient zeroizing unit. An image compression apparatus configured to include a variable length coding unit that allocates more information to data having a higher occurrence probability among the quantized coefficient data. 3. A process for reducing the amount of image data by thinning out each image data by dividing the image data into half by a low-pass filter and a high-pass filter by a predetermined number of steps. Wavelet transforming step for obtaining coefficient data by a wavelet transforming step, a quantizing step for quantizing coefficient data output from the wavelet transforming section for each sub-band to obtain quantized coefficient data, and a wavelet for transforming the quantized coefficient data. Dequantization step to restore the coefficient data of the transform, wavelet inverse transform step to restore the coefficient data restored in the dequantization step to image data, input the restored image data and the current image data, The restored image data is divided into a plurality of blocks composed of a plurality of images, and the block internal structure of the restored image is divided. A motion compensation step of extracting, in each of the divided blocks, vertical and horizontal distances at which the difference between a pixel and a constituent pixel in the block of the current image data is minimized, and performing forward interframe predictive coding based on the difference data. And the held motion scalar amount and the motion scalar amount that is the distance data in the vertical and horizontal directions in the motion compensation unit are compared, and the motion scalar amount that is the distance data in the vertical and horizontal directions is a predetermined amount. When the value exceeds, the quantization coefficient 0 setting step for setting the quantization coefficient data to 0 for the corresponding image area, and the variable length coding step for assigning more information to the data having a higher occurrence probability among the quantization coefficient data. An image compression method configured to have. 4. A held motion scalar quantity is compared with a motion scalar quantity which is distance data in the vertical and horizontal directions in the motion compensation section, and a motion scalar quantity which is distance data in the vertical and horizontal directions. Is greater than a predetermined amount, in place of the quantized coefficient zeroing step for setting the quantized coefficient data to 0 for the corresponding image area, the held motion scalar amount and the vertical and horizontal directions in the motion compensator are set. When the amount of motion scalar, which is the distance data in the vertical and horizontal directions, exceeds the predetermined amount for each subband, the corresponding image area is quantized. Conversion coefficient data to 0
The image compression method according to claim 3, further comprising a quantization coefficient 0 conversion step.