JP2004266720A - Image compression device, image compression method, and image compression program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image compression device hardly causing distortion in inter-frame compression and capable of improving a compressibility. <P>SOLUTION: The image compressing device is equipped with a means of dividing the whole image in a frame into blocks of n×m pixels (n, m: natural numbers), a means of finding a difference quantity between data of a block and data of a block of a last frame, a means of outputting one of large, intermediate, and small difference evaluation results according to a result of comparison of the difference quantity with specified thresholds, a data selecting means of selecting and outputting block data divided as output data when the difference evaluation evaluation result shows a large difference, selecting and outputting difference data between the data of the block and the block data of the last frame when an intermediate difference, and outputting no data when a small difference, a means of reversibly encoding and outputting the data outputted by the data selecting means and the difference evaluation result, and a means of putting a new last frame together according to the difference evaluation result. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image compression device for compressing a moving image, an image compression method, and an image compression program.
[0002]
[Prior art]
Since moving image data has a large data amount, storing it without compression causes a reduction in performance in all of the speed and storage efficiency when using data. Therefore, when storing and using image data, it is essential to perform high-efficiency compression. As a method of this high-efficiency compression, a method of continuously compressing still images such as Motion JPEG, which independently compresses images in the time direction, and a compression method of further reducing redundancy by utilizing the temporal relationship of images as in MPEG. A method has been proposed. An image compression method that further reduces redundancy by utilizing the temporal relevance of the image is required to achieve a higher compression ratio.
[0003]
When generating moving image data by a method of compressing using the relevance of images, one frame of data can be restored by itself, and one frame of data is restored using intra-frame compression and data from previous and subsequent frames. It is general to use two types of compression, namely possible inter-frame compression. The inter-frame compression can increase the compression ratio because the redundancy can be compressed not only between the preceding and succeeding pixels on the screen but also in time, but it cannot be restored unless there is a temporally previous or subsequent image. For this reason, in moving image compression, a GOP (Group of Picture) is formed as a set of compressed data in one frame and a plurality of inter-frame compressed data as a set. MPEG is famous as a compression method combining such intra-frame compression and inter-frame compression.
[0004]
Here, conventional inter-frame compression will be described using MPEG2 as an example. The operation of MPEG2 inter-frame compression will be described below.
(1) A video signal is read into a frame memory and cut out into blocks.
(2) For each block, motion prediction is performed on the data of the frame at the immediately preceding time, and the difference between the data at the corresponding positions at the preceding and following times is calculated.
(3) The difference is encoded by a discrete cosine transform (DCT).
(4) Quantize the DCT code data and output the subband component to the low resolution layer and the high resolution component to the enhancement layer.
(5) Lossless encoding such as entropy encoding and run-length encoding is performed on the predicted motion vector obtained earlier and the quantized data to obtain output data.
(6) The reconstructed data is stored in the frame memory through the inverse quantization and the inverse DCT transform of the output of (4), and is used for frame prediction at the next time.
Note that MPEG1 and MPEG2 are described in Non-Patent Document 1. Further, a technique relating to improvement of MPEG is described in Patent Document 1.
[0005]
[Patent Document 1]
Japanese Patent Application Laid-Open No. H4-177992 [Non-Patent Document 1]
"Bit Separate Volume: Information Compression Technology for Digital Broadcasting and the Internet," Kyoritsu Shuppan, February 5, 1998, p. 128 to p. 145, p. 152-p. 177, p. 205-p. 237
[0006]
[Problems to be solved by the invention]
In the conventional frame-to-frame compression, when a sudden scene change or an object that was not visible until now suddenly appears on the screen, or when it is hidden from the screen, the difference calculation from the screen that is most similar is performed. Is Even in this case, although the process has the closest data, the process is performed between unrelated data, which does not lead to a reduction in redundancy. Conversely, by calculating the difference, a pattern included in the subtracted image that is not in the original image is added, and the restored image may include a portion such as an afterimage. In the case of only a scene change or the like, selection of a compensation image is easy because the average value of the two images is close and the difference or the sum of squares of the difference is small.
On the other hand, when data in which only the DC component which is the average of the target pixel is subtracted from the data to be compressed (no processing in the time direction is performed) is considered, the noise caused by the encoding in a scene change or the like is considered. Is data that is unlikely to occur.
On the other hand, if the compression without the difference calculation is always used, an afterimage due to the difference between the irrelevant image data described above is not generated. However, in this method, the redundancy that has been reduced by the motion prediction compensation for the portion that can be correctly predicted by the motion prediction cannot be reduced, and the data amount in most of the image increases.
[0007]
The present invention has been made in view of such circumstances, and by switching to a compression method in accordance with the relevance of moving image data, distortion is less likely to occur in inter-frame compression and image compression can be improved. It is an object to provide an apparatus, an image compression method, and an image compression program.
[0008]
[Means for Solving the Problems]
The invention according to claim 1 is an image compression apparatus for compressing moving image data between frames, and a block dividing unit that divides an entire image in the frame into blocks of nxm pixels (n and m are natural numbers), A difference calculating means for calculating a difference between the data of the block and the block data of the immediately preceding frame; and a difference evaluation of any of a large difference, a medium difference, and a small difference based on a result of comparing the difference with a predetermined threshold value. A difference evaluation unit that outputs a result, and selects and outputs the block data cut out by the block dividing unit as output data when the difference evaluation result is a large difference, and outputs the block data when the difference is in the middle. Data selecting means for selecting and outputting difference data between the data and the block data of the immediately preceding frame, and selecting no data if the difference is small; A lossless encoding unit that losslessly encodes and outputs the data output by the selection unit and the difference evaluation result, and a data synthesis unit that synthesizes a new immediately preceding frame based on the difference evaluation result. I do.
[0009]
According to a second aspect of the present invention, in the image compression apparatus, the image compression apparatus further includes a block encoding unit that performs block encoding on output data of the data selection unit, and a quantization unit that quantizes output data of the block encoding unit. It is characterized by having.
[0010]
According to a third aspect of the present invention, in the image compression apparatus, a block encoding unit that performs block encoding on data of a block cut out by the block dividing unit, and a quantization unit that quantizes output data of the block encoding unit. And further comprising:
[0011]
According to a fourth aspect of the present invention, there is provided an image compression method for compressing moving image data between frames, wherein a whole image in the frame is divided into blocks of n × m pixels (n and m are natural numbers); A difference calculating step of calculating a difference amount between the block data and the block data of the immediately preceding frame; and a difference evaluation of any of a large difference, a medium difference, and a small difference based on a result of comparing the difference amount with a predetermined threshold value. A difference evaluation step of outputting a result, and selecting and outputting the block data cut out by the block division step as output data when the difference evaluation result is a large difference; A data selection step of selecting and outputting difference data between the data and the block data of the immediately preceding frame, and selecting no data if the difference is small; A lossless encoding step of losslessly encoding the data output in the selection step and the difference evaluation result and outputting the result, and a data synthesis step of synthesizing a new immediately preceding frame based on the difference evaluation result. .
[0012]
According to a fifth aspect of the present invention, there is provided an image compression program for compressing moving image data between frames, wherein the whole image in the frame is divided into blocks of n × m pixels (n and m are natural numbers); A difference calculation process for calculating a difference amount between the data of the block and the block data of the immediately preceding frame, and a difference evaluation of one of a large difference, a medium difference, and a small difference based on a result of comparing the difference amount with a predetermined threshold. A difference evaluation process for outputting a result, and selecting and outputting the block data cut out by the block division process as output data when the difference evaluation result is a large difference; Data selection processing for selecting and outputting difference data between the data and the block data of the immediately preceding frame, and selecting no data if the difference is small; The lossless encoding process that losslessly encodes the data output in the data selection process and the difference evaluation result and outputs the data, and the data synthesis process that synthesizes a new immediately preceding frame based on the difference evaluation result is performed by the computer. It is characterized by the following.
[0013]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an image compression apparatus according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of the embodiment. In this figure, reference numeral 1 denotes a frame memory for sampling and quantizing a video input signal and storing one frame of image data. Reference numeral 2 denotes a block cutout unit that cuts out a block of image data of n × m (n and m are natural numbers) pixels. Reference numeral 3 denotes a difference evaluation unit that evaluates the magnitude of the difference from the immediately preceding frame. Reference numeral 4 denotes a data selection unit that selects one of a plurality of data according to the evaluation result of the difference evaluation unit 3. Reference numeral 5 denotes a combining data selection unit that selects data for creating the immediately preceding frame. Reference numeral 6 denotes a frame memory for storing the immediately preceding frame. Reference numeral 7 denotes a motion compensation prediction unit that performs motion compensation prediction. Reference numerals 8 and 9 are lossless encoding units that losslessly encode input data and output the encoded data. A multiplexing unit that outputs compressed data by multiplexing the encoded data output from the lossless encoding units 8 and 9. Reference numeral 11 denotes a difference output unit that outputs a difference between two image data. Reference numeral 12 denotes a combining unit that combines two pieces of image data.
[0014]
Next, an image compression operation by inter-frame compression of the image compression apparatus shown in FIG. 1 will be described with reference to FIG.
First, the block cutout unit 2 cuts out block data of n × m pixels from the frame memory 1. The block data cut out here is output to the motion compensation prediction unit 7. In response to this, the motion compensation prediction unit 7 searches the immediately preceding frame stored in the frame memory 6 for one block having image data closest to the block to be processed. And outputs the result to the difference output unit 11. As a condition of the closest data, a difference between pixel values at each point of the block is obtained, and one block at a place where the sum of absolute values is the smallest is selected. Alternatively, instead of calculating the motion assuming that there is no motion, a block at the same position in the immediately preceding frame may be always selected.
[0015]
Next, the difference output unit 11 obtains a difference (a difference between values of pixels) for each pixel between the motion compensation data and the block data cut out by the block cutout unit 2. Then, the absolute values of the differences of the pixels are summed, and this sum is referred to as a difference amount. The difference output unit 11 outputs the obtained difference amount to the difference evaluation unit 3.
[0016]
Alternatively, after calculating the difference for each pixel of the block data obtained by cutting out the difference amount of the pixel block, the variance of the difference may be calculated and used as the difference amount. This is a calculation method that considers the case where the screen is uniformly brightened. If the screen is uniformly brightened, when the difference is calculated with the pixels of the highly relevant block, the difference is centered around 0. And the error is the value ± error of the brightened portion. At this time, the error is minimized by minimizing the variance of the difference instead of the sum of the absolute values of the difference. This makes it possible to provide a compression method having a wide application range in which the compression ratio does not decrease even when the light source changes.
[0017]
In response, the difference evaluation unit 3 evaluates the difference amount. If the difference amount exceeds the threshold Th1, the difference evaluation unit 3 determines that no relevant compensation data has been found, and notifies the data selection unit 4 of the result of the difference evaluation as "large difference". When the difference amount does not exceed the threshold value Th1 but exceeds the threshold value Th2 (where Th1>Th2> 0), it is determined that there is a change relevant to the immediately preceding frame, and the result of the difference evaluation is set to “ Notifying the data selection unit 4 of "differential". If the difference amount does not exceed Th2, it is determined that the previous frame and the inside of the block have not changed, and the result of the difference evaluation is notified to the data selection unit 4 as "small difference".
[0018]
Next, the data selection unit 4 selects output data based on the difference evaluation result notified from the difference evaluation unit 3. When the difference evaluation result is “large difference”, the data selection unit 4 selects the block data cut out by the block cutout unit 2 and outputs the block data as it is. In the case of “difference in difference”, block data (output of the difference output unit 11) is selected by subtracting the pixel of each point of the motion compensation data from the pixel of each point of the target block data, and this block data is Output. If the difference is “small difference”, the information is replaced with the no-block information and output.
[0019]
Next, the lossless encoding unit 9 losslessly encodes the data selected and output by the data selection unit 4. The encoding method used here performs Huffman encoding, which is one of entropy encoding, and outputs the result.
In parallel with this, the lossless encoding unit 8 performs Huffman encoding of information (for example, large difference = 1, medium difference = 2, small difference = 3) which can identify the difference evaluation result output from the difference evaluation unit 3. Is used for lossless encoding.
Note that a suitable coding such as an arithmetic coding and a range code method may be used as the entropy coding.
[0020]
Next, the multiplexing unit 10 multiplexes and outputs the outputs of the two lossless encoding units 8 and 9. This is compressed data obtained by compressing one block data.
[0021]
Next, the synthesizing unit 12 re-synthesizes the data, and the synthesizing data selecting unit 5 selects the data based on the output of the difference evaluating unit 3 and outputs the selected data to the frame memory 6. That is, in the case of "large difference", the block of the output of the data selection unit 4 is added, and in the case of the difference, the block obtained by adding the value of the immediately preceding frame (after motion compensation) to the output data of the data selection unit 4 If the difference is small, the data of the immediately preceding frame (after motion compensation) is output to the frame memory 6.
By executing the above processing for all blocks, compressed data of an image is output.
[0022]
It should be noted that the two thresholds Th1 and Th2 described above may be changed, and adjustments may be made to increase the image quality to lower the compression ratio or lower the compression ratio to increase the image quality as needed.
[0023]
Next, another embodiment will be described with reference to FIG. The device shown in FIG. 2 is different from the device shown in FIG. 1 in that a block coding unit 13, a quantization unit 14, and a block decoding unit 15 are newly provided. That is, quantization is performed. Since the output of the data selection unit 4 can take from -255 to 255, the value is normalized to 0 to 255. The block encoding unit 13 shows features / components specific to an image from a set (block) of fixed-size rectangular pixels, such as discrete cosine transform (or other orthogonal transform), BTC (Block Truncation Coding), and vector quantization. A process for extracting a value and expressing it as a small amount of data is performed. As encoded data output when the discrete cosine transform is used, n × m frequency components (Qij) are output as a DCT sequence, and the average of all pixels is output as a DC component. In addition, the quantization unit 14 obtains and multiplies a constant that falls within the range of −128 to +127 after multiplying Qij which is originally a real number, and performs integer conversion according to a certain rule such as truncation to reduce the absolute value. Under a certain condition (for example, a value of 2 or less, after a zigzag scan is performed and 0 is found at least once), a process closer to 0 is further performed. The block decoding unit 15 decodes the block data encoded by the block encoding unit 13 to combine the data.
[0024]
Next, another embodiment will be described with reference to FIG. The device shown in FIG. 3 differs from the device shown in FIG. 2 in that a block coding unit 13 and a quantization unit 14 are provided between the block cutout unit 2 and the difference output unit 11. The operations of the block encoding unit 13 and the quantizing unit 14 shown in FIG. 3 are the same as those shown in FIG.
[0025]
Next, a case where BTC (Block Trunking Coding) is used in the block encoding unit 13 shown in FIGS. Japanese Patent Application Laid-Open No. 60-87596 describes a compression method using BTC in the case of RGB expression, and further describes a usage method in Japanese Patent No. 3260910. When color is block-coded by the BTC method, YUV420 is used instead of RGB. The BTC process will be briefly described below. First, the average of all the pixels in the block is calculated and set as an average gradation value M. Subsequently, the average value M is subtracted from each pixel in the block, and the average value R of each is calculated as a gradation amplitude index R. Pixels of M or more are represented by gradation number 1 and pixels of M or less are represented by gradation number 0, and a sequence of gradation numbers is output as a vector having n × m elements. Thus, encoding can be performed.
[0026]
Next, in the apparatus shown in FIG. 3, block coding of block data is performed using BTC (Block Trunking Coding), and a method of calculating a difference amount will be described. The output of the BTC is the average color information, the average gradation value indicating the undulation from the average, and the quantization value (for the number of pixels) indicating which gradation belongs to each pixel.
The difference amount is: difference amount = difference in average color information × k1 (average color coefficient) + difference in color tone index × k2 (tone value coefficient) + | difference in quantization value | × (average of average tone value ). Here, | difference in quantization value | is a value obtained by calculating the Euclidean distance between two vectors, considering the quantization value as an n × m-dimensional vector. Further, k1 and k2 are weighting coefficients, and the same value is used for all moving images.
[0027]
Next, a method of using vector quantization in the block encoding unit 13 shown in FIGS. Compression by vector quantization is described in JP-A-64-60059. Japanese Patent Application Laid-Open No. Sho 60-146363 discloses vector quantization with motion compensation. The operation when vector quantization is applied to block coding will be briefly described.
First, before the vector quantization, the average Eb of the block and the maximum amplitude Rb from the average are calculated, and the variation from the average is normalized to (eg, −128 to +128) using Eb and Rb. Then, for a block of n × m pixels of the normalized fluctuation value, a vector having an element configuration close to a vector having the values of the pixels constituting the block as elements is selected from a prepared vector table, and the index number Ib of the vector is selected. And Eb and Rb are the results of quantizing the block. When selecting a vector, the sum of squares of the differences between the elements is calculated so that the one with a smaller Euclidean distance becomes the closer vector. By this operation, a vector quantization process can be performed.
[0028]
Next, an operation of performing block encoding by vector quantization and calculating a difference amount in the apparatus illustrated in FIG. 3 will be described. First, before the vector quantization, the average Eb of the block and the maximum amplitude Rb from the average are calculated, and the variation from the average is normalized to -128 to +128 using Eb and Rb. Then, a vector having the closest value to the normalized variation is searched for, and its index number Ib, Eb, and Rb are taken as the result of quantizing the block.
The difference is calculated by the following equation: Difference = Eb difference × K1 (average color coefficient) + Rb difference × K2 (amplitude coefficient) + average of difference | × Rb of individual elements of two vectors indicated by Ib. Here, | ... | is assumed to be a vector having a difference value in the n × m dimension as an element, and a value obtained by calculating a Euclidean distance between the two vectors. Further, k1 and k2 are weighting coefficients, and constant values are used in all compression in the moving image.
[0029]
Next, the operation of the apparatus shown in FIG. 3 in which the discrete cosine transform is used for block coding and the difference amount calculation is performed using the DC component DC which is the output after the orthogonal transform and the element Qij after the transform is described.
The amount of change is obtained by the amount of change = DC difference + Σ (| Qij of target image−Qij of comparative image × Kij). Here, Kij (i <n, j <m) is a weighting coefficient, and a constant value is used in all compressions in the moving image, and K is designed to satisfy Kij ≧ Ki + 1j and Kij ≧ Kij + 1. The equation shown here is a weighted sum of the differences, and the weights are weighted so that the low-frequency components to which human vision reacts sensitively become large so that the values become close to human senses. Thus, a more natural restored image can be obtained.
[0030]
It should be noted that attribute values generated by encoding block data (corresponding to an average gradation value, gradation amplitude, and quantization value (sequence of gradation numbers) in BTC, and DC components and AC frequency components in DCT). Corresponding to the corresponding Qij), or each of the independent attributes is grouped to obtain a difference, and the difference is evaluated based on the difference for each attribute value. Furthermore, the difference, the original value, and the data are calculated for each attribute. You may make it select none. For example, when the BTC method is used for block coding, the average grayscale value and the grayscale amplitude are independent, and if the average grayscale value is the same and the grayscale amplitude changes greatly, the grayscale amplitude and grayscale By transmitting only the vector of the number and not transmitting the average gradation value, by using the same value as before at the time of decoding, the same image can be reproduced with less information transmission. According to this, for example, when the pattern changes or deforms without changing the background color, the data amount decreases. Even when only the reverse average gradation value changes, it can be reproduced with a small amount of information transmission. According to this, for example, when the entire screen becomes evenly bright due to a change in the light source, the data amount is reduced. As described above, for the independent elements such as the average gradation value and the gradation amplitude, the difference amount is separately evaluated for large, medium, and small, and the loss amount is composed of the lossless encoding using the original image block and the difference value respectively corresponding to each. By transmitting only necessary data by reversible coding using blocks and no data, the amount of data is reduced and the compression rate is improved.
[0031]
As described above, in inter-frame compression, a frame is divided into blocks, and a time change is inspected. In the middle of the same information, compressed data using the relationship between frames is transmitted in the middle, and the compression method is switched according to the relationship of data, so distortion is hardly generated by inter-frame compression and the compression ratio Can be improved.
[0032]
In addition, a program for realizing the function of each processing unit in FIGS. 1 to 3 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read and executed by a computer system. Compression processing may be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer system” also includes a WWW system provided with a homepage providing environment (or display environment). The “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, and a CD-ROM, and a storage device such as a hard disk built in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) inside a computer system serving as a server or a client when a program is transmitted through a network such as the Internet or a communication line such as a telephone line. In addition, programs that hold programs for a certain period of time are also included.
[0033]
Further, the above program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the "transmission medium" for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. Further, the program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.
[0034]
【The invention's effect】
As described above, according to the present invention, it is proposed to perform encoding of an original image without subtracting compensation by a predicted image when the relevance is small, and to perform compensation by a predicted image when the relevance is large. Thus, afterimage noise can be reduced while maintaining a high compression ratio. Also, as a method of checking the relevancy, the difference between the predicted image data and the original image data is used. Since the simple determination that the relevance is large is made, the determination that the data amount is increased in descending order of the necessity can be performed at high speed. In particular, when the difference is smaller than a predetermined threshold value, the fluctuation is often noise or fluctuation, and is a change that is hidden by the noise or fluctuation and cannot be discriminated by humans. Therefore, it is determined that there is no difference, and data is generated. By not doing so, it becomes possible to reduce the amount of transmission data as compared with the method of always sending differential data or compressed data within a frame. Also, the present invention can be applied to a compression method having a feature of dividing image frame data into blocks and performing encoding and quantization in blocks. As the block encoding method, discrete cosine transform, BlockTruncationCoding, vector quantization, or the like is used. It becomes possible.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention.
FIG. 2 is a block diagram showing a configuration of another embodiment of the present invention.
FIG. 3 is a block diagram showing a configuration of another embodiment of the present invention.
[Explanation of symbols]
1, 6 ... frame memory 2 ... block cutout unit 3 ... difference evaluation unit 4 ... data selection unit 5 ... synthesis data selection unit 7 ... motion compensation prediction unit 8, 9, ... ..Reversible encoding unit 10 Multiplexing unit 11 Difference output unit 12 Synthesizing unit 13 Block encoding unit 14 Quantizing unit

Claims (5)

An image compression apparatus for compressing moving image data between frames, comprising:
Block dividing means for dividing the entire image in the frame into blocks of n × m pixels (n and m are natural numbers);
Difference calculation means for calculating the difference between the data of the block and the block data of the immediately preceding frame,
A difference evaluation unit that outputs a difference evaluation result of any of a large difference, a difference, and a small difference based on a result of comparing the difference amount and a predetermined threshold value;
When the difference evaluation result is a large difference, the block data cut out by the block dividing means is selected and output as output data, and when the difference is being calculated, the data of the block and the block data of the immediately preceding frame are compared. Data selection means for selecting and outputting the difference data and not selecting the data when the difference is small;
Lossless encoding means for outputting the data output by the data selection means and the difference evaluation result in a lossless manner,
An image compression device comprising: a data synthesizing unit that synthesizes a new immediately preceding frame based on the difference evaluation result. The image compression device, block encoding means for block encoding the output data of the data selection means,
The image compression apparatus according to claim 1, further comprising a quantization unit that quantizes output data of the block encoding unit. The image compression device, block encoding means for block encoding the data of the block cut out by the block dividing means,
The image compression apparatus according to claim 1, further comprising a quantization unit that quantizes output data of the block encoding unit. An image compression method for compressing moving image data between frames,
A block division process of dividing the entire image in the frame into blocks of n × m pixels (n and m are natural numbers);
A difference calculation step of calculating a difference amount between the data of the block and the block data of the immediately preceding frame,
A difference evaluation step of outputting a difference evaluation result of one of a large difference, a difference, and a small difference based on a result of comparing the difference amount with a predetermined threshold;
When the difference evaluation result is a large difference, the block data cut out by the block division process is selected and output as output data, and when the difference is in the middle, the data of the block is compared with the block data of the immediately preceding frame. A data selection step of selecting and outputting the difference data and not selecting the data when the difference is small;
A lossless encoding step of losslessly encoding and outputting the data and the difference evaluation result output in the data selection step,
A data synthesizing step of synthesizing a new immediately preceding frame based on the difference evaluation result. An image compression program for compressing moving image data between frames,
A block division process of dividing the entire image in the frame into blocks of nxm pixels (n and m are natural numbers);
A difference calculation process for calculating a difference amount between the data of the block and the block data of the immediately preceding frame;
Based on the result of comparing the amount of difference with a predetermined threshold, a difference evaluation process of outputting a difference evaluation result of any of a large difference, a difference, and a small difference,
When the difference evaluation result is a large difference, the block data cut out by the block division processing is selected and output as output data, and when the difference is being calculated, the data of the block and the block data of the immediately preceding frame are compared. A data selection process for selecting and outputting the difference data and not selecting the data when the difference is small;
A lossless encoding process in which the data output in the data selection process and the difference evaluation result are losslessly encoded and output,
An image compression program for causing a computer to perform a data synthesis process for synthesizing a new immediately preceding frame based on the difference evaluation result.

Images