CN100450181C - An embedded image compression technique based on wavelet transformation - Google Patents
An embedded image compression technique based on wavelet transformation Download PDFInfo
- Publication number
- CN100450181C CN100450181C CNB031344216A CN03134421A CN100450181C CN 100450181 C CN100450181 C CN 100450181C CN B031344216 A CNB031344216 A CN B031344216A CN 03134421 A CN03134421 A CN 03134421A CN 100450181 C CN100450181 C CN 100450181C
- Authority
- CN
- China
- Prior art keywords
- coding
- wavelet transformation
- amplitude
- coefficient
- bit
- 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.)
- Expired - Fee Related
Links
Images
Abstract
The present invention relates to an embedded image compression technique on the basis of wavelet transformation, which has the technical thought that because coefficients after eight scale type wavelet transformation have various favorable easy compression properties, such as favorable empty frequency localization properties, frequency domain energy aggregating properties, energy attenuation properties, space aggregating properties of high frequency coefficients, etc., eight scale type wavelet transformation is firstly carried out to input images. The present invention is characterized in that subbands after wavelet transformation are divided into low frequency portions and high frequency portions according to wavelet transformation properties, low frequency subbands are independently compressed in a predictive coding mode, high frequency subbands are encoded in a bit plane mode, and then respective results are merged to form code flows. The present invention has the advantages of low complexity, little calculated amount, little occupied storage space and small influence of image statistics properties, and the present invention is particularly suitable for images with complex texture. Therefore, the present invention can be applied to conditions in a large amount, such as hardware realization, etc., which have strict complexity requirements, and important economic effect is exerted.
Description
Affiliated technical field: the present invention relates to a kind of embedded image compress technique, mainly any natural image or artificial image are compressed, realize progressive transmission with extremely low complexity based on wavelet transformation.
Background technology:
The compression of image and video is to influence the transmission rate of picture signal and the key issue of quality in the communication.In compression standard PEG, MPEG series, H.26x serial, comprise in the up-to-date still image compression standard JPEG 2000, all used a kind of run length encoding method of classics.Just a data sequence is represented with Run/Level that wherein Run represents in the sequence number of null value between two nonzero values, Level represents the value of next nonzero coefficient, then the Run/Level symbol sebolic addressing is used the Huffman encoding technology.In order to improve the real-time capacity of coding and decoding, be easy to hardware and realize that standard has been set up a fixing Huffman code table according to the statistics to great amount of images simultaneously.Only need according to the inquiry in corresponding code table of Run/Level symbol just passable during coding and decoding.The number that a shortcoming clearly of this processing method is exactly the Run symbol can't determine in advance, and generally has manyly, and the Level symbol equally also has this problem.And, more can bring huge symbol expense with Run/Level symbol combined coding, make the decrease in efficiency of whole entropy coding.
The video compression standard of a new generation H.26L in, changed traditionally with the method for Run/Level symbol to combined coding, Run and Level are encoded respectively.But still can not well solve the problem of big symbol weight.
In more current popular wavelet compression methods, all use arithmetic coding to improve the efficient of entropy coding, improve the deficiency of Huffman encoding.Arithmetic coding is from whole symbol sebolic addressing, adopts recursive form to carry out continuous programming code.There is not one-to-one relationship between incoming symbol and the code word.Can approach the entropy of information source by the mark bit, having broken through in the Huffman encoding each symbol can only be by the restriction of an integer bit near the entropy of information source, by transmitting the symbol of big probability to reach compression effects with still less mark bit.When the symbol sebolic addressing of needs codings increases, the stream rate that the utilization arithmetic coding obtains can be slowly near the entropy of information source.
Though all used arithmetic coding, organize the strategy of coefficient different, bring the difference of complexity and efficient.For example based on the method for zero tree, EZW, SPIHT, SFQ search repeatedly to each layer wavelet coefficient during owing to compression, and whether the existence of searching zero tree, so need expend a large amount of operation time and memory space.And based on the rate distortion estimation theory, to the method that coded bit stream carries out reprocessing, for example SFQ is in order to judge the contribution of each zero tree, influence to the rate distortion curve, and quantization step is to zero tree result's influence, and the rate distortion that need carry out repeatedly judges that complexity is the highest; EBCOT is in order to compare the distortion performance of each code block, also need to carry out the data that the Tier-2 process makes up each code block, this performance to its method has played critical effect, we can say that the performance raising of these class methods is that cost obtains to sacrifice operation time.For utilizing the expand method of classification wavelet coefficient of morphology, MRWD, SLCCA, though improved compression performance, all utilization in various degree the process that expands of interband, so just improved the complexity of compressing.
Summary of the invention:
For avoiding the defective of prior art, the present invention proposes a kind of embedded image compress technique based on wavelet transformation.According to encode coefficient behind the wavelet transformation of the mode of bit-planes, eliminate coding to Level, and with Run with its binary form arithmetic coding, thereby solve the shortcoming of the huge symbol weight of Run/Level, the shortcoming that Huffman encoding efficient is not high.Because this method is not utilized the interband correlation, just write down, compressed the position of wavelet coefficient in the band, according to the order of each subband zig-zag to whole bit-planes single treatment, without any repeatable operation, so in Embedded image compression algorithm, complexity is minimum.The image compression result who utilizes this method to realize is gratifying, and performance is robust relatively, goes for the image of any kind.
Technological thought of the present invention is: because the coefficient behind the wavelet transformation of octave stepwise has the multiple good characteristics that are easy to compress such as space clustering characteristic of good empty localization property, frequency domain energy accumulating and Energy Decay, high frequency coefficient frequently.So the image of input is carried out the wavelet transformation of octave stepwise earlier, it is characterized in that: according to the characteristic of wavelet transformation, subband behind the wavelet transformation is divided into low frequency part and HFS, and low frequency sub-band is carried out the independent compression of predictive coding formula, high-frequency sub-band is carried out the coding of bit-planes formula, and the result merges the formation code stream separately then.
The range value of low frequency sub-band coefficient is generally all bigger, has occupied more than 90% of whole transformation energy, and bigger deviation is arranged, so can be the low frequency sub-band independent compression.Because the correlation in the low frequency sub-band between pixel is very high, the low frequency sub-band proximate coefficients has bigger correlation than high-frequency sub-band.Therefore more effective with the predictive coding meeting, also more reasonable.In this case,, only handle the difference of prediction pixel and original pixel, reduce the deviation of coefficient with neighborhood pixel prediction present picture element, can the more efficient compression coefficient, the raising compression ratio.Concrete steps are: to the wavelet low frequency subband, with neighborhood pixel prediction present picture element, carry out the prediction based on the neighborhood tonsure; Then to the amplitude of prediction difference and symbol individual processing respectively: the symbol of prediction difference is carried out symbolic coding based on context separately, and amplitude is carried out adaptive arithmetic amplitude coding according to bit-planes.
The distribution of high-frequency sub-band coefficient is the laplacian distribution of needle pattern, and less coefficient has bigger amplitude.But because natural image is ever-changing, the position of these coefficients is random distribution, can't learn in advance.Though high frequency coefficient has only the sub-fraction of whole energy, this part coefficient has mainly been described the marginal portion information of original image, and the definition of reconstructed image is had significant effects.And having space clustering, just important coefficient is distributed near the edge of original image mostly, and in other local one-tenth sparse distribution.So can utilize this characteristic high frequency coefficient is carried out Run-Length Coding.
On each bit-planes of high frequency coefficient there are two steps, importance checking process and amplitude optimizing process.The importance checking process is the coefficient compression to importance the unknown, and these coefficients are handled with adaptive Run-Length Coding, and distance of swimming value is with its binary form arithmetic coding.Because adopt bit-planes to quantize wavelet coefficient, on so each bit-planes, all coefficients are 0 or 1.So the Level value of this moment must be 1, need not encode, and only use its symbol use is handled based on the symbolic coding of context.The known coefficient of amplitude optimizing process compression importance, because importance is known, just positional information obtains, and only need express the value that this coefficient presented and get final product on current bit-planes, so use amplitude optimization encoding process.
High frequency coefficient is carried out Run-Length Coding: owing to adopt bit-planes to quantize wavelet coefficient, on so each bit-planes, all coefficients are 0 or 1.So the Level value of this moment must be 1, need not encode, use symbolic coding processing to get final product its symbol and only use based on context.For any one distance of swimming value Run, big or small change at random can't be determined at all in advance.If yet be converted into binary representation, can regard as by limited 0 and 1 and form.It is b that Run is converted to binary system
kb
K-1... b
1b
0(k ∈ N), wherein b
kBe the most significant bit (MSB), b
0It is the bit (LSB) of meaning least.So just represent distance of swimming value, replace the original distance of swimming is worth metric coding with 0, the 1 binary sequence mode of forming.When actual treatment, only need the corresponding binary value 0 or 1 of coding to get final product.
For the distance of swimming value of non-zero, its MSB must be 1, so though a total k+1 position in fact only needs coding k position to get final product, need not encode to its MSB position.And distance of swimming value must be non-negative, so without coded identification.In order to increase the length of the distance of swimming, the distance of swimming need not be confined to can cross over the subband border in the subband, writes down the coefficient situation of next height band in addition.
Here one have 3 symbols in Bian Ma the model,, also need an END symbol to represent the end-of-encode of current distance of swimming value except 0,1 two outer symbol.Because if with numerical value of binary coding, decoder can't determine when end-of-encode,, show that the coding of this distance of swimming value this moment is finished so need special symbol to notify decoder.Adaptive arithmetic code device one has 3 symbols like this.The flow process of specific coding distance of swimming value is as follows, and decoding just need add 1 in highest order in contrast, just can obtain actual distance of swimming value.
For?i＝k-1?to?0
If?b
i＝1
Encode?1
Else
Encode?0
Encode?symbol?END
Adaptive cataloged procedure, solved the big symbol weight of conventional method, the shortcoming that efficient is not high, whole cataloged procedure does not need extra internal memory and priori, do not need to store in advance any Huffman code table yet, fully can be according to the adaptive coding of the statistical property of input picture.In the statistics distance of swimming, can consider the restriction of subband in addition, increase run length, improve code efficiency.The process of method is that the distance of swimming of statistics is described with its binary form, only the binary character that coding is corresponding gets final product, and in order to indicate finishing of a distance of swimming, adds an END end mark, set up the adaptive arithmetic code of 3 models, distance of swimming value is compressed.
Amplitude optimization coding: meaningful if some coefficient shows in last once bit-plane coding, again these coefficients are carried out the importance inspection with regard to not needing in the bit-plane coding so afterwards, and judge directly whether this coefficient is still meaningful in the scanning of this bit-planes.This amplitude optimizing process does not need to consider positional information, can effectively utilize former result.The amplitude optimizing process of some wavelet compression algorithms in the past, for example EZW, SPIHT etc., its amplitude optimizing process be in each bit-planes scanning, is identical for the treating method of each coefficient.Yet in fact, this part also can according to the contextual information of neighborhood and information compressed classified.Experiment shows that amplitude optimizes bit and the compressed peculiar weak correlation of ratio of current coefficient, with the amplitude of neighborhood pixel weak correlation is arranged also.Like this, just can set up corresponding context model during compression and instruct compression according to correlation a little less than this.
Description of drawings:
Fig. 1: the theory diagram of embedded image compress technique
Fig. 2: the schematic diagram of the wavelet transformation of octave stepwise
Fig. 3: be respectively " Barbara " image, " Lena " image, " Mandrill " image
Fig. 4: the R-D curve of test pattern
Solid line is the Barbara image, and dotted line is the Lena image, and chain-dotted line is the Mandrill image
Embodiment:
Now in conjunction with the accompanying drawings the present invention is further described:
Fig. 1 is the theory diagram of embedded image compress technique, and Fig. 2 is the schematic diagram that 512 * 512 standard testing image " Lena " is carried out the wavelet transformation of octave stepwise.And these characteristics can well instruct compression as priori.So the first step of the present invention is the wavelet transformation that received image signal is carried out the octave stepwise.Characteristic according to conversion is divided into low frequency part and HFS with coefficient then, adopts different processing methods respectively.
Through behind the wavelet transformation, very big difference has appearred in the low frequency part of subband and the statistical property of HFS.Low frequency sub-band presents and the similar statistical property of original image, and high-frequency sub-band and original image present very big difference.Low frequency sub-band is carried out the independent compression of predictive coding formula, high-frequency sub-band is carried out the Run-Length Coding of bit-planes formula, the result merges the formation code stream separately then.The low frequency sub-band benefit of coding separately is even without the information of receiving any high-frequency sub-band, by with the low frequency sub-band interpolation, also can produce coarse original image, watches for the recipient.
High-frequency sub-band is carried out in the Run-Length Coding of bit-planes formula, adopt the adaptive run-length encryption algorithm, the benefit that this technology is brought is: 1, can represent different distance of swimming values effectively with less symbol: though the difference between two decimal numbers may be very big, but be converted into binary system, length may only have slight difference, can not bring very big variation, for example 4 and 15 differ 11, length only differs from 1 when being expressed as binary system, and 4 and 31,63, length only differs from 2 between 127,3,4, but value differs from 27,59,123, just only need the seldom several symbols of many volumes just can represent a very big numerical value, and a very big numerical value only just can be represented with several symbols seldom.So this adaptive run-length encryption algorithm is for the long sequence of distance of swimming value, just the more sequence of null value is well suited for.2, without any need for priori, fully according to the adaptive coding of picture material: compare with traditional Run-Length Coding, the adaptive run-length encryption algorithm does not here need extra internal memory and priori, do not need to store in advance any Huffman code table yet, fully can be according to the adaptive coding of the statistical property of input picture.3, the cross-border statistics of run length: because coding and decoding has identical scanning sequency,, increased run length, raised the efficiency so in the statistics distance of swimming, can consider the restriction of subband.
Fig. 4 tests the gray level image (as Fig. 3, being respectively " Barbara " image, " Lena " image, " Mandrill " image) of 3 standard testings of 512 * 512.Use 9/7 biorthogonal wavelet, decompose 5 layers, use symmetric extension.Distortion performance uses the Y-PSNR (PSNR) that often occurs in the literature as criteria of quality evaluation, is defined as:
Wherein f (m, n) and
Be respectively (m, n) gray value of corresponding pixel in the gray value of individual pixel and the reconstructed image in the image.The R-D curve of any as can be seen from Figure 3 width of cloth test pattern is unusual flat-satin all, chattering do not occur.This is by the decision of the characteristics of algorithm, and whole algorithm is carried out sequential scanning according to bit-planes and carried out, and can recover with best quality at any point of cut-off, so sudden change can not occur.The another one characteristics are that the image for complex texture has stronger robustness, for example to " Barbara " image and " Mandrill " image, do not have too big variation slope.
In sum, the present invention has following characteristics:
● wavelet transform provides a tight multiresolution characteristic of image;
● account for the preferentially compression fully of low frequency component of the most energy of conversion coefficient;
● successive approximation to quantification provides a tight multiple-accuracy representing of significant coefficient, and simplifies embedded algorithm;
● based on the symbolic coding of context can the more efficient use wavelet conversion coefficient the symbol correlation;
● adaptive Run-Length Coding can be with less symbolic representation than long distance of swimming value;
● the arithmetic coding of self adaptation multi-model provides symbol sebolic addressing entropy coding method fast and effectively, and it does not need training and stores any code table in advance;
● the arithmetic coding of the multi-model of the amplitude optimizing process optimization bit of more effectively classifying;
● when reaching the target bit rate or the distortion factor, coding and decoding just stops immediately, can accurately reach the bit rate of expection.
As the above analysis, complexity of the present invention is extremely low, and amount of calculation is little, and the memory space that takies is few.Be subjected to the influence of image statistics characteristic less, be particularly suited for the image of complex texture.Method with the binary form direct coding distance of swimming can combine with the algorithm of other any use Run-Length Codings, improves the coding effect.So the present invention is expected to be widely used in the occasions strict to complexity such as hardware realization, has given play to great economic benefits.
Claims (6)
1, a niche is in the Embedded Image Coding of wavelet transformation, input picture is carried out the wavelet transformation of octave stepwise, it is characterized in that: according to the characteristic of wavelet transformation, subband behind the wavelet transformation is divided into low frequency part and HFS, and low frequency sub-band is carried out the independent compression of predictive coding formula, high-frequency sub-band is carried out the Run-Length Coding of bit-planes formula, and the result merges the formation code stream separately then; The independent compression of described predictive coding formula for low frequency sub-band is: to the wavelet low frequency subband, with neighborhood pixel prediction present picture element, carry out the prediction based on the neighborhood tonsure; Then to the amplitude of prediction difference and symbol individual processing respectively: the symbol of prediction difference is carried out symbolic coding based on context separately, and amplitude is carried out adaptive arithmetic amplitude coding according to bit-planes.
2, a kind of Embedded Image Coding according to claim 1 based on wavelet transformation, it is characterized in that: the bit-planes formula Run-Length Coding to high-frequency sub-band is: earlier the high-frequency sub-band of wavelet transformation is carried out bit-planes and quantize, two steps were arranged on each bit-planes, importance checking process and amplitude optimizing process: be divided into three kinds of processing:, these coefficients are handled with adaptive Run-Length Coding to the coefficient compression of importance the unknown; The symbol use is handled based on the symbolic coding of context; The coefficient use that importance is known is encoded based on the amplitude optimization of context.
3, a kind of Embedded Image Coding based on wavelet transformation according to claim 2 is characterized in that: self adaptation rider coding: any one distance of swimming value Run is converted to binary system b
kb
K-1... b
1b
0Expression, wherein: k ∈ N, b
kBe the most significant bit MSB, b
0Be the bit LSB of meaning least, represent distance of swimming value, replace the original distance of swimming is worth metric coding, only need the corresponding binary value 0 or 1 of coding to get final product by the 0 and 1 limited binary sequence mode of forming.
4, a kind of Embedded Image Coding according to claim 3 based on wavelet transformation, it is characterized in that: adaptive arithmetic code device one has 3 symbols, except 0,1 two outer symbol, also need an END symbol to represent the binary sequence end-of-encode of current distance of swimming value correspondence.
5, a kind of Embedded Image Coding based on wavelet transformation according to claim 4 is characterized in that: for the distance of swimming value of non-zero, its MSB=1 only needs coding k position to get final product, and need not encode in the MSB position.
6, a kind of Embedded Image Coding according to claim 2 based on wavelet transformation, it is characterized in that: amplitude optimization coding: optimize the bit ratio peculiar weak correlation compressed with current coefficient according to amplitude, with the amplitude of neighborhood pixel weak correlation is arranged also, set up corresponding context model and instruct compression, use 3 models of 2 symbols that 3 context models of 2 symbols set up to be: model 1: the corresponding coefficient that had had amplitude to optimize; Model 2: the corresponding coefficient that did not also have amplitude to optimize is important but have a coefficient at least in 8 pixels in 3 * 3 neighborhoods around it; Model 3: the corresponding coefficient that did not also have amplitude to optimize, but around it in 8 pixels in 3 * 3 neighborhoods the neither one coefficient be important, just this pixel is the important pixel that isolates; Described model 2 and 3 all is the coefficient that corresponding amplitude is for the first time optimized, but need select corresponding arithmetic coding model according to the significance state of neighborhood.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031344216A CN100450181C (en) | 2003-07-23 | 2003-07-23 | An embedded image compression technique based on wavelet transformation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB031344216A CN100450181C (en) | 2003-07-23 | 2003-07-23 | An embedded image compression technique based on wavelet transformation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1571514A CN1571514A (en) | 2005-01-26 |
| CN100450181C true CN100450181C (en) | 2009-01-07 |
Family
ID=34470192
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB031344216A Expired - Fee Related CN100450181C (en) | 2003-07-23 | 2003-07-23 | An embedded image compression technique based on wavelet transformation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100450181C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104125449A (en) * | 2014-08-12 | 2014-10-29 | 上海众恒信息产业股份有限公司 | Image processing system and method |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101354151B1 (en) * | 2006-08-24 | 2014-01-28 | 삼성전자주식회사 | Method and apparatus for transforming and inverse-transforming image |
| CN101631243B (en) * | 2009-08-07 | 2011-10-26 | 清华大学 | Image encoding/decoding method based on wavelet transformation |
| CN102006483B (en) * | 2009-09-03 | 2013-05-01 | 中兴通讯股份有限公司 | Video coding and decoding method and device |
| CN101984666A (en) * | 2010-11-19 | 2011-03-09 | 南京邮电大学 | Image lossless compression and decompression method based on lifting wavelet transform |
| CN102222350A (en) * | 2011-05-31 | 2011-10-19 | 上海交通大学 | Image compressing device and method based on wavelet transformation and arithmetic coding |
| CN112422992A (en) * | 2020-10-28 | 2021-02-26 | 西安万像电子科技有限公司 | Image coding method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6272180B1 (en) * | 1997-11-21 | 2001-08-07 | Sharp Laboratories Of America, Inc. | Compression and decompression of reference frames in a video decoder |
| WO2002035849A1 (en) * | 2000-10-24 | 2002-05-02 | Eyeball Networks Inc. | Three-dimensional wavelet-based scalable video compression |
| US20020080408A1 (en) * | 1999-12-17 | 2002-06-27 | Budge Scott E. | Method for image coding by rate-distortion adaptive zerotree-based residual vector quantization and system for effecting same |
| US6477280B1 (en) * | 1999-03-26 | 2002-11-05 | Microsoft Corporation | Lossless adaptive encoding of finite alphabet data |
| US6519284B1 (en) * | 1999-07-20 | 2003-02-11 | Koninklijke Philips Electronics N.V. | Encoding method for the compression of a video sequence |
-
2003
- 2003-07-23 CN CNB031344216A patent/CN100450181C/en not_active Expired - Fee Related
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6272180B1 (en) * | 1997-11-21 | 2001-08-07 | Sharp Laboratories Of America, Inc. | Compression and decompression of reference frames in a video decoder |
| US6477280B1 (en) * | 1999-03-26 | 2002-11-05 | Microsoft Corporation | Lossless adaptive encoding of finite alphabet data |
| US6519284B1 (en) * | 1999-07-20 | 2003-02-11 | Koninklijke Philips Electronics N.V. | Encoding method for the compression of a video sequence |
| US20020080408A1 (en) * | 1999-12-17 | 2002-06-27 | Budge Scott E. | Method for image coding by rate-distortion adaptive zerotree-based residual vector quantization and system for effecting same |
| WO2002035849A1 (en) * | 2000-10-24 | 2002-05-02 | Eyeball Networks Inc. | Three-dimensional wavelet-based scalable video compression |
Non-Patent Citations (2)
| Title |
|---|
| JPEG2000第一部分最终委员会草案1.0版（JPEG 2000 Part I Final Committee Draft Version 1.0）. 联合二值图像专家组（Joint Bi-level Image Experts Group）、联合图像专家组(Joint Photographic Experts Group),附录D第3章第3节. 2000 |
| JPEG2000第一部分最终委员会草案1.0版（JPEG 2000 Part I Final Committee Draft Version 1.0）. 联合二值图像专家组（Joint Bi-level Image Experts Group）、联合图像专家组(Joint Photographic Experts Group),附录D第3章第3节. 2000 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104125449A (en) * | 2014-08-12 | 2014-10-29 | 上海众恒信息产业股份有限公司 | Image processing system and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1571514A (en) | 2005-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hsiang | Embedded image coding using zeroblocks of subband/wavelet coefficients and context modeling | |
| Tang et al. | Three-dimensional wavelet-based compression of hyperspectral images | |
| US6671413B1 (en) | Embedded and efficient low-complexity hierarchical image coder and corresponding methods therefor | |
| CN1322442A (en) | Encoding method for compression of video sequence | |
| Ponomarenko et al. | High-quality DCT-based image compression using partition schemes | |
| CN101631243B (en) | Image encoding/decoding method based on wavelet transformation | |
| CN101132530A (en) | Method for implementing built-in image compression based on run-length coding | |
| CN100450181C (en) | An embedded image compression technique based on wavelet transformation | |
| Bhammar et al. | Survey of various image compression techniques | |
| Chen et al. | Image coding based on wavelet transform and uniform scalar dead zone quantizer | |
| T Hashim et al. | Color image compression using DPCM with DCT, DWT and quadtree coding scheme | |
| Abdul-Wahed et al. | Compression of image using multi-wavelet techniques | |
| Chai et al. | Statistically adaptive wavelet image coding | |
| CN100477789C (en) | A method of inter-frame reinforcement based on Moving Joint Picture Experts Group 2000 | |
| CN110365990A (en) | A kind of quasi- lossless video encoding system in narrowband | |
| Wu et al. | Lossy to lossless compressions of hyperspectral images using three-dimensional set partitioning algorithm | |
| Hsin | Texture segmentation in the joint photographic expert group 2000 domain | |
| Berghom et al. | Context conditioning and run-length coding for hybrid, embedded progressive image coding | |
| Ranjeeta et al. | Image compression: an overview | |
| Hsin et al. | Adaptive binary arithmetic coder-based image feature and segmentation in the compressed domain | |
| Oliver et al. | A new fast lower-tree wavelet image encoder | |
| Muzaffar et al. | Linked significant tree wavelet-based image compression | |
| Karthikeyan et al. | A survey on embedded zero tree wavelet | |
| Chen et al. | LBT based low complexity image compression method | |
| Hsin et al. | Image segmentation in the JPEG2000 domain |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| ASS | Succession or assignment of patent right |
Owner name: XIBEIYICHUAN MULTI-MEDIA INFORMATION TECHNOLOGY CO Free format text: FORMER OWNER: NORTHWESTERN POLYTECHNICAL UNIVERSITY Effective date: 20091030 |
|
| C41 | Transfer of patent application or patent right or utility model | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20091030 Address after: Room 9, Times Building, No. 3, No. 402, hi tech Zone, Xi'an, Shaanxi Patentee after: Northwest Yichuan Multimedia Information Technology Co. Ltd. Address before: No. 127 Youyi West Road, Shaanxi, Xi'an Patentee before: Northwestern Polytechnical University |
|
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090107 Termination date: 20100723 |