CN109040759B - Image parallel compression device and method - Google Patents
Image parallel compression device and method Download PDFInfo
- Publication number
- CN109040759B CN109040759B CN201810839894.1A CN201810839894A CN109040759B CN 109040759 B CN109040759 B CN 109040759B CN 201810839894 A CN201810839894 A CN 201810839894A CN 109040759 B CN109040759 B CN 109040759B
- Authority
- CN
- China
- Prior art keywords
- module
- quantization
- value
- coding
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/42—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation
- H04N19/436—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals characterised by implementation details or hardware specially adapted for video compression or decompression, e.g. dedicated software implementation using parallelised computational arrangements
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/124—Quantisation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
Abstract
The invention relates to a device and a method for image parallel compression, which evenly divide image data into 5 times of blocks, and after a clock comes, one pixel of an m-th image block is sent to an m/5-th remainder receiving module; the predictor module combines the original pixel value X and the predicted pixel value
Sending the data to an error calculation module; an error calculation module for calculating the original pixel value X and the predicted pixel value
The difference value of the difference value is sent to a quantization module; the quantization module calculates | Rc‑Rb|+|Rc‑RaAnd quantizing and sending to a coding module; the encoding module encodes the quantized value and sends the encoded value to the overflow processing module; the output processing module judges whether overflow occurs in the quantization process, and if the overflow does not occur, codes are output; and outputting the code and the overflow code if the overflow occurs. The remote sensing image parallel compression method based on the flowing water has the characteristics of low hardware cost, high processing speed, strong applicability, easy engineering realization, stability and reliability, and the design method has obvious advantages in the application aspect of the aerospace field.
Description
Technical Field
The invention belongs to the technical field of satellite remote sensing, and relates to a parallel compression device and method based on running water.
Background
With the development of space-to-earth observation technology, the amount of data required to be transmitted on the satellite is increased sharply, and higher requirements are put on the compression capability of satellite images. Complex algorithm implementation and limited hardware resources become main factors influencing the on-satellite compression processing capacity, and complex video compression algorithms such as static compression algorithms JEPG-LS, JPEG2000, particularly H.265 and the like put higher requirements on the processing capacity during hardware implementation.
At present, static image compression technologies used in the satellite remote sensing field include JEPG-LS and JPEG2000, which are affected by application environment and algorithm complexity, and have not little limitation on processing capacity and efficiency, and the ASIC mode solves the limitation of the programmable chip, but correspondingly increases manufacturing cost. Dynamic image compression techniques such as h.264, especially h.265, have higher algorithm complexity, and it is not easy to solve the contradiction between processing power and algorithm complexity.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the parallel compression technology overcomes the defect of insufficient processing capacity of the existing compression technology, provides a remote sensing image parallel compression technology based on the running water, has extremely low computational complexity while providing a compression function, adopts extremely few resources and has extremely high processing capacity.
The above object of the present invention is achieved by the following technical solutions:
the image parallel compression device comprises an image segmentation module, five receiving modules, a predictor module, an error calculation module, a quantization module, a coding module and an output processing module;
and the image segmentation module is used for uniformly dividing the image data into k image blocks, wherein k is a multiple of 5. After the clock comes, one pixel of the mth image block is sent to the m/5 th receiving modules of the compression device; adding 1 to m in each clock cycle, switching m-5 to the next five image block sequences after all pixels of the five image blocks are transmitted every 5 clock cycles, and adding m +5 to the next five image block sequences until all the pixels of the image blocks are transmitted; k is more than or equal to m and more than or equal to 1;
the five receiving modules send the pixels of the received image block to the predictor module; the predictor module receives one pixel per clock cycle and predicts the received pixel, the next clock cycle combines the original pixel value X with the predicted pixelValue of
Sending the data to an error calculation module; an error calculation module for calculating the original pixel value X and the predicted pixel value
And the original pixel value X and the predicted pixel value are compared in the next clock cycle
The calculated difference value is sent to a quantization module; the quantization module calculates | Rc-Rb|+|Rc-RaL, and quantizing, and sending the quantized value and the received difference value to the encoding module in the next clock cycle, Ra,Rb,RcPixels representing left, upper, and upper left side points, respectively, of the received pixel; the coding module codes the quantized value and sends the difference value between the code and the received value to the overflow processing module in the next clock period; the output processing module judges whether overflow occurs in the quantization process, and if the overflow does not occur, the code is output; outputting the code and an overflow code if an overflow occurs.
Preferably, the method for acquiring the overflow coding comprises: judging whether the received difference exceeds a threshold value, and if not, outputting the codes and a plurality of fixed bits of 0; if the difference exceeds the part of the truncation range, the equal interval quantization is carried out, and the equal interval quantization value is directly coded.
Preferably, the predictor module prediction model is as follows:
preferably, the quantization module transforms | Rc-Rb|+|Rc-RaAnd comparing the | with a threshold value, if the | is larger than the threshold value, adopting a long quantization table, and otherwise, adopting a short quantization table.
Preferably, the first and second liquid crystal materials are,the quantization module also performs a pixel reconstruction,
sending the output by an output processing module through a coding module, wherein qErr is | Rc-Rb|+|Rc-RaAnd (5) calculating the result of | according to the calculation result.
Preferably, the coding module uses a long coding table to code if the quantized value is greater than a set threshold, otherwise uses a short coding table to code.
Preferably, the display device further comprises a top module for outputting a clock signal and receiving externally transmitted image data.
A method for performing image parallel compression by using the image parallel compression device comprises the following steps:
(1) the image segmentation module is used for uniformly dividing the image data into k image blocks; let i equal to 1, m equal to 1;
(2) the image segmentation module sends ith pixels of the m image blocks to the m/5 th parallel receiving modules after a clock arrives; adding 1 to m every clock cycle, and adding m to 5 every 5 clock cycles;
(3) the predictor module receives one pixel per clock cycle and predicts the received pixel, the next clock cycle will be the original pixel value X and the predicted pixel value
Sending the data to an error calculation module;
(4) an error calculation module for calculating the original pixel value X and the predicted pixel value
And the original pixel value X and the predicted pixel value are compared in the next clock cycle
The calculated difference value is sent to a quantization module;
(5) the quantization module calculates | Rc-Rb|+|Rc-RaI, go togetherLine quantization, in the next clock cycle, the quantized value and the received difference value are sent to an encoding module;
(6) the coding module codes the quantized value and sends the difference value between the code and the received value to the overflow processing module in the next clock period;
(7) the output processing module judges whether overflow occurs in the quantization process, and if the overflow does not occur, the code is output; if overflow occurs, judging whether the received difference value exceeds a threshold value, if not, outputting the code and fixed bits of several 0; if the difference exceeds the part of the truncation range, carrying out equal interval quantization, and directly coding the equal interval quantization value;
(8) judging whether the pixel coding is finished, if so, entering the step (9), otherwise, returning to the step (2) by adding i + 1;
(9) judging whether the coding of all image blocks is finished or not, and finishing the parallel compression of the image if the coding is finished; if not, returning m +5, i to 1 to the step (2);
preferably, the quantization module in step (5) converts | Rc-Rb|+|Rc-RaAnd comparing the | with a threshold, if the | is larger than the threshold, quantizing by adopting a long quantization table, and otherwise quantizing by adopting a short quantization table.
Preferably, in the step (6), if the quantized value is greater than the set threshold, the coding module uses a long coding table for coding, otherwise, uses a short coding table for coding.
Preferably, the method also comprises a step (10) of converting the five compressed code streams according to a required format and formatting a protocol after the parallel image pipeline processing is completed.
Compared with the prior art, the invention has the following advantages:
(1) the parallel compression method adopts a maximum 5-path parallel compression mode, can greatly improve the real-time processing rate of data compared with a single-path compression mode, improves the data processing rate by 5 times, can meet the task requirement of high-speed data transmission of the current remote sensing image, and conforms to the development trend of real-time data processing.
(2) The parallel compression method is suitable for the existing linear array and area array images, can set the size of the image block at will, and is more flexible compared with the existing compression method.
(3) The parallel compression method of the invention does not need to change the format of the front input parallel data before parallel compression, and can directly call the compression module to complete the parallel coding processing. The processing flow is simple and easy to realize, and the complexity of using the encoder is greatly reduced.
(4) The parallel compression method of the invention does not need to buffer the data, reduces the error probability in the data processing process, avoids the phenomena of number loss, dislocation and the like in the process of using FIFO/RAM to buffer the data, and avoids the defect of retaining partial data in the buffer.
(5) The parallel compression method only needs one clock frequency in the whole compression process, avoids the problem of clock domain crossing caused by data serial-parallel change and a plurality of clock frequencies introduced in the shunting process, and reduces the risk of a metastable state and unnecessary clock resource waste.
(6) The parallel compression method saves resources, and compared with the existing compression mode, the method can reduce the usage amount of FIFO resources and the usage amount of logic resources.
Drawings
FIG. 1 is a flow chart of the compression algorithm of the present invention;
FIG. 2 is a flow chart of the design of the pipelined parallel compression method of the present invention;
FIG. 3 is a schematic diagram of a pipelined parallel compression method according to the present invention;
FIG. 4 is a diagram of a pipelined parallel compression method predictor model according to the present invention.
Detailed Description
With reference to fig. 3, the image parallel compression apparatus of the present invention includes a top module, an image segmentation module, five receiving modules, a predictor module, an error calculation module, a quantization module, a coding module and an output processing module;
the image segmentation module is used for uniformly dividing image data into k image blocks, wherein k is a multiple of 5, and after a clock comes, one pixel of the mth image block is sent to the m/5 th remainder receiving module; adding 1 to m in each clock cycle, switching m-5 to the next five image blocks after all pixels of the five image blocks are sent every 5 clock cycles, and adding m +5 until all the image blocks are sent; k is more than or equal to m and more than or equal to 1;
and the five receiving modules send the pixels of the received image block to the FIFO for reading and storing, and send the pixels to the predictor module, one receiving module sends the pixels every one clock cycle, the next cycle is switched to the next receiving module for sending, and the five cycles are circulated once.
The predictor module receives one pixel per clock cycle and predicts the received pixel, the next clock cycle will be the original pixel X and the predicted pixel
Sending the data to an error calculation module; the prediction process has a variety of mature predictors, such as MED predictor of JPEG-LS. Using the relationship of adjacent pixels, as shown in fig. 4, the predicted value is:
an error calculation module for calculating the original pixel X and the predicted pixel
And the original pixel X, the predicted pixel are processed in the next clock cycle
The calculated difference value is sent to a quantization module;
the quantization module reads the stored pixels from the FIFO and calculates | Rc-Rb|+|Rc-RaL, and quantizing, and sending the quantized value and the received difference value to the encoding module in the next clock cycle, Ra,Rb,RcRespectively representing the left of the received pixelPixels of side, upper, and upper left side points; and a rom buffer of the quantization module is a long quantization table and a short quantization table respectively. According to the difference (| R) of the upper, left and upper left pixels of the pixel to be codedc-Rb|+|Rc-Ra|) selecting a quantization table for quantization to obtain a quantization difference value qErr, and if the difference value is greater than a certain fixed value, adopting a long quantization table; otherwise, a short quantization table is adopted. Using quantized difference qErr and predicted value
Reconstruction pixel
Stored for subsequent prediction. Because the original quantization table and the coding table occupy more storage resources in the FPGA and have a large number of repeated quantization values, parameters stored in the rom are optimized according to the characteristics of the coding table at EX _ adjust, and the resource overhead of the rom is reduced.
The encoding module encodes the quantized value and sends the encoded difference value and the received difference value to the overflow processing module in the next clock period, and a rom of the encoding module stores a long Huffman encoding table and a short Huffman encoding table; the output processing module judges whether overflow occurs in the quantization process, and if the overflow does not occur, the code is output; outputting the code and an overflow code if an overflow occurs. Since the single-path code stream needs to be independently organized, 5 paths of instantiations are needed for the module. The module carries out code stream splicing according to the input single code, and carries out zero filling operation according to the final code stream output result.
The top interface module inputs reset, clock, 5-way gate control and data, the width and height of the image block and outputs code stream enable, data, end identification and code stream length. The code stream enabling comprises an identifier for outputting the code stream to be effective, a gating signal and a pull-up signal, wherein the pull-up represents that the code stream is effective. The aspect of the image block may be adapted according to the specific application.
A remote sensing image parallel compression method based on flow, which is combined with figures 1-3, the concrete realization method comprises the following steps:
(1) and (3) partitioning the CCD camera image data with standard or non-standard width into multiple image blocks of 5.
(2) The image segmentation module sends the ith pixel of the mth image block to the m/5 th parallel receiving modules after the clock arrives; adding 1 to m every clock cycle, and adding m to 5 every 5 clock cycles;
(3) the predictor module receives one pixel per clock cycle and predicts the received pixel, the next clock cycle will be the original pixel value X and the predicted pixel value
Sending the data to an error calculation module; the predictor predicts the pixel of the ith pixel point to be coded of the mth image block, and the predicted pixel value is counted as
The prediction process has various mature predictors, such as MED predictor of JPEG-LS, and by using the relation of adjacent pixels and combining with the graph of FIG. 4, the predicted values are as follows:
wherein R isa,Rb,RcRespectively representing pixel values of points to the left, upper, and upper left of the ith point.
(4) An error calculation module for calculating the original pixel value X and the predicted pixel value
And the original pixel value X and the predicted pixel value are compared in the next clock cycle
The calculated difference value is sent to a quantization module; the error calculation module calculates the original pixel value X and the predicted pixel value of the ith pixel point 1 of the mth image block
The difference of (a):
predicting the pixel of the ith point to be coded of the (m + 1) th image block by a predictor module in the same clock cycle;
(5) the quantization module calculates | Rc-Rb|+|Rc-RaQuantizing, and sending the quantized value and the received difference value to a coding module in the next clock period; for the mth image block 1, according to the difference (| R) of the upper, left and upper left pixels of the ith pixel point to be codedc-Rb|+|Rc-Ra|) selecting a quantization table for quantization to obtain a quantization difference value qErr, and if the difference value is greater than a certain fixed value, adopting a long quantization table; otherwise, a short quantization table is adopted. Using quantized difference qErr and predicted value
Reconstruction pixel
And storing the pixel parameters for subsequent prediction as pixel parameters of adjacent lines of the subsequent prediction. The first row of pixels and the first pixel of the second row are directly output without coding processing. The error calculation module in the same clock cycle processes the (m + 1) th image block 2, and the predictor module predicts the (m + 3) th image block; the long quantization table and the short quantization table select a compression multiple according to the probability distribution of the prediction residual value and design a segmentation interval and a quantization value;
the compression multiple determines the value and the segmentation interval of the long and short quantization tables, and the quantization value and the segmentation interval under the fixed compression ratio are found out through training of a large number of images. All the quantization parameters are selected by adopting a model training method, each compression ratio corresponds to one set of parameters, namely if 2 compression ratios are set, 2 sets of quantization tables and coding tables are correspondingly set. For example, when 10-bit quantization is performed by 2-fold compression, the quantization range of the long quantization table is [ -152,152], the quantization range of the short quantization table is [ -32,32], and the possible prediction residual range is [ -1023, 1023], and the truncation process is performed without full quantization. Meanwhile, more than 2 times of quantization numbers are adopted for the long quantization table, the quantization number is 36(2 times of compression is 5bit quantization, namely 32 quantization intervals); while the short quantization table has a quantization number of less than 32 quantization intervals (28).
(6) The coding module codes the quantized value and sends the difference value between the code and the received value to the overflow processing module in the next clock period; and performing Huffman coding on the quantized value of the mth image block 1, if the quantized value is greater than a set threshold, adopting a long coding table for coding, and otherwise, adopting a short coding table for coding. Meanwhile, the quantization module processes the (k + 1) th image block 2, the error calculation module processes the (k + 2) th image block 3, and the predictor module predicts the (k + 1) th image block 4; as image detail information is reserved and a preset compression multiple is achieved, a long quantization table and a short quantization table are selected according to the neighborhood flatness, and a segmentation interval and a quantization value are designed according to the probability distribution of a prediction residual value and a target compression multiple. The short quantization table reserves more detailed information for the image, the long quantization table has more requirements on adjustment of the compression code rate, and the two quantization tables are matched for use, so that the corresponding image quality is ensured while the target compression ratio is achieved. The long coding table and the short coding are used for training long and short quantization values according to a large number of images, the long and short quantization values are respectively subjected to Huffman coding to generate corresponding coding tables, and the coding is performed in a table look-up mode through the long and short coding tables, so that hardware resources are saved, and the Huffman coding is not required to be completed in hardware.
(7) The output processing module judges whether overflow occurs in the quantization process, and if the overflow does not occur, the code is output; if overflow occurs, judging whether the received difference value exceeds a threshold value, if overflow occurs, judging whether a residual error Err exceeds the threshold value, and if not, outputting Huffman coding results and a plurality of 0 fixed bits; if the Err exceeds the threshold, the overflow processing is carried out on the image block 1, because the truncation processing is adopted for quantization, equal-interval quantization is carried out on the part of the residual error exceeding the truncation range, the Huffman coding is directly carried out on the quantized value, and the Huffman coding of the quantized value of the image block 1 and the Huffman coding of the part of the residual error exceeding the truncation range are output. In order to distinguish the code streams of the truncation quantization and the overflow quantization, the overflow quantization code words of the fixed bit must be connected after the code words at two ends of the truncation quantization are specified, and when the overflow is not generated, the overflow code words are 0 s of the fixed bit. The coding module in the same clock cycle processes the (m + 1) th image block 2, the quantization module processes the (m + 2) th image block 3, the error calculation module processes the (m + 3) th image block 4, and the predictor module predicts the (m + 4) th image block 5;
(8) judging whether the pixel coding is finished, if so, entering the step (9), otherwise, returning to the step (2) by adding i + 1;
(9) judging whether the coding of all image blocks is finished, if so, finishing the parallel pipeline processing of the images, and then performing the step (10); if not, returning m +5, i to 1 to the step (2);
(10) and the parallel multi-path compressed code streams are converted and formatted according to a required format.
The above description is only for the best mode of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention should be covered within the scope of the present invention.
Those skilled in the art will appreciate that the invention may be practiced without these specific details.
Claims (6)
1. The image parallel compression device is characterized by comprising an image segmentation module, five receiving modules, a predictor module, an error calculation module, a quantization module, a coding module and an output processing module;
the image segmentation module is used for uniformly dividing the image data into k image blocks, wherein k is a multiple of 5; after the clock comes, one pixel of the mth image block is sent to the m/5 th receiving modules of the compression device; adding 1 to m in each clock cycle, switching m-5 to the next five image block sequences after all pixels of the five image blocks are transmitted every 5 clock cycles, and adding m +5 to the next five image block sequences until all the pixels of the image blocks are transmitted; k is more than or equal to m and more than or equal to 1;
the five receiving modules send the pixels of the received image block to the predictor module;the predictor module receives and predicts a pixel per clock cycle, the next clock cycle being the period of the original pixel value X and the predicted pixel value
Sending the data to an error calculation module; an error calculation module for calculating the original pixel value X and the predicted pixel value
And sending the calculated difference to the quantization module in the next clock cycle; the quantization module calculates | Rc-Rb|+|Rc-RaL, and quantizing, and transmitting the quantized value and the received difference value to the encoding module in the next clock cycle, Ra,Rb,RcPixels representing left, upper, and upper left side points, respectively, of the received pixel; the coding module codes the quantization value and sends the code and the received difference value to the output processing module in the next clock period; the output processing module judges whether overflow occurs in the quantization process, and if the overflow does not occur, the code is output; outputting the code and an overflow code if an overflow occurs;
judging whether the received difference exceeds a threshold value, and if not, outputting the codes and a plurality of fixed bits of 0; if the difference exceeds the threshold value, carrying out equal interval quantization on the part of the difference exceeding the truncation range, and directly coding the equal interval quantization value;
the predictor module predicts the model as follows:
the quantization module transforms | Rc-Rb|+|Rc-RaComparing | with a threshold, if the | is larger than the threshold, adopting a long quantization table, otherwise, adopting a short quantization table;
the quantization module also performs a pixel reconstruction,
sending the output by an output processing module through a coding module, wherein qErr is | Rc-Rb|+|Rc-RaThe calculated result of | is;
and if the quantized value of the coding module is larger than the set threshold value, the coding module codes by adopting a long coding table, otherwise, the coding module codes by adopting a short coding table.
2. The apparatus of claim 1, further comprising a top module outputting a clock signal for receiving the image data transmitted from the outside.
3. A method for parallel compressing an image by using the apparatus of claim 1, comprising the steps of:
(1) the image segmentation module is used for uniformly dividing the image data into k image blocks; let i equal to 1, m equal to 1;
(2) the image segmentation module sends ith pixels of the m image blocks to the m/5 th parallel receiving modules of the compression device after a clock arrives; adding 1 to m every clock cycle, and adding m to 5 every 5 clock cycles;
(3) the predictor module receives and predicts a pixel per clock cycle, the next clock cycle being the period of the original pixel value X and the predicted pixel value
Sending the data to an error calculation module;
(4) an error calculation module for calculating the original pixel value X and the predicted pixel value
And sending the calculated difference to the quantization module in the next clock cycle;
(5) the quantization module calculates | Rc-Rb|+|Rc-RaAnd quantized, at the next clock cycleThe quantized value and the received difference value are sent to a coding module;
(6) the coding module codes the quantized value and sends the difference value between the code and the received value to the output processing module in the next clock period;
(7) the output processing module judges whether overflow occurs in the quantization process, and if the overflow does not occur, the code is output; if overflow occurs, judging whether the received difference value exceeds a threshold value, if not, outputting the code and fixed bits of several 0; if the difference exceeds the part of the truncation range, carrying out equal interval quantization, and directly coding the equal interval quantization value;
(8) judging whether the pixel coding is finished, if so, entering the step (9), otherwise, enabling i +1, and returning to the step (2);
(9) judging whether the coding of all image blocks is finished or not, and finishing the parallel compression of the image if the coding is finished; if not, returning m +5, i to 1 to the step (2).
4. The method of parallel image compression as claimed in claim 3, wherein the quantization module in step (5) transforms | Rc-Rb|+|Rc-RaAnd comparing the | with a threshold, if the | is larger than the threshold, quantizing by adopting a long quantization table, and otherwise quantizing by adopting a short quantization table.
5. The method of parallel image compression as claimed in claim 3, wherein in step (6), the coding module uses a long coding table for coding if the quantized value is greater than the set threshold, otherwise uses a short coding table for coding.
6. The method of parallel image compression as claimed in claim 3, further comprising a step (10) of converting and protocol formatting the five compressed code streams according to a required format after the parallel image pipeline processing is completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810839894.1A CN109040759B (en) | 2018-07-27 | 2018-07-27 | Image parallel compression device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810839894.1A CN109040759B (en) | 2018-07-27 | 2018-07-27 | Image parallel compression device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109040759A CN109040759A (en) | 2018-12-18 |
| CN109040759B true CN109040759B (en) | 2021-11-16 |
Family
ID=64646883
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810839894.1A Active CN109040759B (en) | 2018-07-27 | 2018-07-27 | Image parallel compression device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109040759B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110324632A (en) * | 2019-05-29 | 2019-10-11 | 西安空间无线电技术研究所 | A kind of data processing and verification method based on OpenMP multi-core parallel concurrent mechanism |
| CN112437308A (en) * | 2020-11-12 | 2021-03-02 | 北京深维科技有限公司 | WebP coding method and device |
| CN112839230B (en) * | 2021-01-20 | 2023-04-28 | 上海思朗科技有限公司 | Image data compression device and image data decompression device |
| CN113068037B (en) * | 2021-03-17 | 2022-12-06 | 上海哔哩哔哩科技有限公司 | Method, apparatus, device, and medium for sample adaptive compensation |
| CN113780549A (en) * | 2021-08-05 | 2021-12-10 | 鹏城实验室 | Quantitative model training method, device, medium and terminal equipment for overflow perception |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101276471A (en) * | 2008-05-12 | 2008-10-01 | 北京中星微电子有限公司 | Method and device for compression of ADPCM image |
| WO2010029850A1 (en) * | 2008-09-09 | 2010-03-18 | 日本電気株式会社 | Image coding device, image decoding device, image coding method, image decoding method, and program |
| CN103246499A (en) * | 2013-05-28 | 2013-08-14 | 西安电子科技大学 | Device and method for parallelly processing images |
| CN103533351A (en) * | 2013-09-26 | 2014-01-22 | 西安空间无线电技术研究所 | Image compression method for multi-quantization table |
| CN105227959A (en) * | 2014-07-03 | 2016-01-06 | 深圳中兴力维技术有限公司 | For odd encoder point shuffling flowing water method and the device thereof of Video coding |
| CN107071450A (en) * | 2016-02-10 | 2017-08-18 | 同济大学 | Coding, coding/decoding method and the device of data compression |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101394560B (en) * | 2008-06-30 | 2010-09-01 | 浙江大学 | Mixed production line apparatus used for video encoding |
| JP2014007469A (en) * | 2012-06-21 | 2014-01-16 | Canon Inc | Image coding device and image coding method |
-
2018
- 2018-07-27 CN CN201810839894.1A patent/CN109040759B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101276471A (en) * | 2008-05-12 | 2008-10-01 | 北京中星微电子有限公司 | Method and device for compression of ADPCM image |
| WO2010029850A1 (en) * | 2008-09-09 | 2010-03-18 | 日本電気株式会社 | Image coding device, image decoding device, image coding method, image decoding method, and program |
| CN103246499A (en) * | 2013-05-28 | 2013-08-14 | 西安电子科技大学 | Device and method for parallelly processing images |
| CN103533351A (en) * | 2013-09-26 | 2014-01-22 | 西安空间无线电技术研究所 | Image compression method for multi-quantization table |
| CN105227959A (en) * | 2014-07-03 | 2016-01-06 | 深圳中兴力维技术有限公司 | For odd encoder point shuffling flowing water method and the device thereof of Video coding |
| CN107071450A (en) * | 2016-02-10 | 2017-08-18 | 同济大学 | Coding, coding/decoding method and the device of data compression |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109040759A (en) | 2018-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109040759B (en) | Image parallel compression device and method | |
| JPS6226633B2 (en) | ||
| US7525456B2 (en) | Methods for adaptively selecting entropy encoding modes | |
| JPS58197984A (en) | Adaptive forecasting encoding device of television signal | |
| CN101438597B (en) | Image data compression device, compression method, and image data decompression device, decompression method | |
| CN108347602B (en) | Method and apparatus for lossless compression of video data | |
| JP2911682B2 (en) | Motion compensation using the minimum number of bits per motion block as a criterion for block matching | |
| CN1953551A (en) | A device and method for image compression | |
| JP2004007555A (en) | Entropy encoder | |
| EP0734166A2 (en) | Apparatus for encoding an image signal having a still object | |
| US9356620B2 (en) | Method for producing video coding and programme-product | |
| AU2001293994B2 (en) | Compression of motion vectors | |
| KR102071459B1 (en) | Encoding apparatus and encoding method | |
| AU2001293994A1 (en) | Compression of motion vectors | |
| CN102273207A (en) | Image conversion device, method, and program | |
| CN111683258B (en) | Image data compression method and interface circuit | |
| JPH09200762A (en) | Method and device for encoding video signal | |
| KR100219133B1 (en) | Method and apparatus for selecting transform coefficient of transform encoding system | |
| CN104995917A (en) | Self-adaption motion estimation method and module thereof | |
| JPH0748859B2 (en) | Predictive coding method for television signals | |
| KR0162201B1 (en) | Dc dpcm system for image data | |
| CN113766238B (en) | Image high-speed compression method and system based on FPGA under JPEG-LS lossless compression standard | |
| CN109040520B (en) | Motion estimation storage architecture design method based on data compression and search range prediction | |
| JP3397682B2 (en) | Image coding device | |
| CN100563344C (en) | The method of self adaptation counter-scanning in adaptive scanning and the decompression in the video compression |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |