CN105072446A - Color image compression sampling and reconstruction algorithm - Google Patents
Color image compression sampling and reconstruction algorithm Download PDFInfo
- Publication number
- CN105072446A CN105072446A CN201510442038.9A CN201510442038A CN105072446A CN 105072446 A CN105072446 A CN 105072446A CN 201510442038 A CN201510442038 A CN 201510442038A CN 105072446 A CN105072446 A CN 105072446A
- Authority
- CN
- China
- Prior art keywords
- passage
- channel
- sample rate
- coloured image
- steps
- 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.)
- Granted
Links
Abstract
The invention relates to a color image compression sampling and reconstruction algorithm. A coding end decomposes a color image into R, G and B channels and respectively carries out wavelet transforming on each channel. For wavelet transforming coefficients of different layers, different sampling rates are used for measuring. The relevance of three channels is used to carry out set sparse reconstruction on the coding end. According to the invention, when the image is processed, less data are sampled, and the restored image quality is great; and compared with a traditional RGB independent reconstruction algorithm, the algorithm provided by the invention has the advantage that the peak signal to noise ratio is improved 1 to 2dB at low sampling rate.
Description
Technical field
The present invention relates to a kind of compression sampling and restructing algorithm of coloured image.
Background technology
Compressed sensing is a kind of novel information processing theory, and it breaches the restriction of nyquist sampling theorem in traditional sampling, makes signal to sample far below nyquist sampling rate, and still can reconstruct primary signal in decoding end high probability.
The application of compressive sensing theory in gray level image of current many scholar's research, and its application in coloured image is comparatively rare.Most scholar is when being applied to coloured image by compressed sensing, adopting transforms on RGB or YUV color space by coloured image, thus being decomposed into three autonomous channels, the compressive sensing theory of recycling gray level image compresses respectively three passages and reconstructs.But adopt and carry out compression reconfiguration coloured image in this way, the high correlation between three passages cannot be utilized, cause the waste of sampling.
Nagesh proposes EJSM algorithm, this utilizes the correlation between three passages to reconstruct for the first time in color image compression perception, article proposes the method for the wavelet coefficient joint sparse of three passages, extract the public part between them, reduce the degree of rarefication of coloured image, thus obtain better quality reconstruction.Majumdar proposes the sparse principle of utilization group to reconstruct coloured image, effectively improves the reconstruction quality of coloured image, but calculation matrix is too large, easily exceedes computer run internal memory in simulations, can only be applicable to undersized coloured image.
Summary of the invention
In view of this, the object of the present invention is to provide a kind of compression sampling and restructing algorithm of coloured image, improve the sampling efficiency of coloured image further, improve reconstruction quality, for the natural image compressed sensing of low sampling rate condition.
For achieving the above object, the present invention adopts following technical scheme: a kind of compression sampling of coloured image and restructing algorithm, is characterized in that:
Programming process comprises the steps:
Steps A 1 a: coloured image is decomposed into R passage, G passage and channel B;
Steps A 2: described R passage, G passage and channel B are carried out three layers of wavelet decomposition respectively and obtains corresponding wavelet conversion coefficient, wherein said R channel decomposition is R1 part, R2 part, R3 part and R4 part; Described G channel decomposition is G1 part, G2 part, G3 part and G4 part; Described channel B is decomposed into B1 part, B2 part, B3 part and B4 part;
Steps A 3: according to the openness distribution sample rate of described wavelet conversion coefficient, carries out the block sampling of described R passage, G passage and channel B;
Decode procedure comprises the following steps:
Step B1: described R1 part, G1 part, B1 part are merged into group1, described R2 part, G2 part, B2 part are merged into group2, described R3 part, G3 part, B3 part are merged into group3, described R4 part, G4 part, B4 part are merged into group4;
Step B2: sparse reconstruct is organized respectively to described group1, group2, group3 and group4;
Step B3: from carrying out four parts isolating three passages group1, group2, group3 and group4 after organizing sparse reconstruct, and resequence, obtains the wavelet coefficient of R passage, G passage and the channel B reconstructed;
Step B4: the picture content that wavelet inverse transformation obtains reconstruct is carried out respectively to described R passage, G passage and channel B;
Step B5: the combination of the picture content of described reconstruct is recovered described coloured image.
Further, being allocated as follows of sample rate in described steps A 3: suppose that described coloured image is N × N coloured image, its sample rate is ratio, in R passage, R1 part accounts for 1/64 of total pixel, make the sample rate ratio1=0.9 of R1 part, in R passage, R2 part accounts for 3/64 of total pixel, make the sample rate ratio2=0.8 of R2 part, in R passage, R3 part accounts for 3/16 of total pixel, R4 part accounts for 3/4 of total pixel, and the sample rate making R3 part is 2 times of R4 fractional-sample rate, i.e. ratio3=2 × ratio4, constant for ensureing the whole-sample rate of image, meet following formula:
Can be obtained fom the above equation:
,
The sample rate method of salary distribution as R passage is performed to described G passage and channel B.
The present invention compared with prior art has following beneficial effect: during process coloured image of the present invention, the data volume of required sampling is few, the picture quality recovered is good, be particularly useful for the natural image compressed sensing of low sampling rate condition, compared to the traditional algorithm that RGB independently reconstructs, when low sampling rate, Y-PSNR about improves 1-2dB.
Accompanying drawing explanation
Fig. 1 is overview flow chart of the present invention.
Fig. 2 is three layers of wavelet decomposition schematic diagram of one embodiment of the invention R passage.
Fig. 3 is the group1 definition procedure figure of one embodiment of the invention.
Fig. 4 is the group2 definition figure of one embodiment of the invention.
Fig. 5 is the group3 definition figure of one embodiment of the invention.
Fig. 6 is the group4 definition figure of one embodiment of the invention.
Fig. 7 is SL20 restructing algorithm flow chart of the present invention.
Embodiment
Below in conjunction with drawings and Examples, the present invention will be further described.
Please refer to Fig. 1, the invention provides a kind of compression sampling and restructing algorithm of coloured image, it is characterized in that:
Programming process comprises the steps:
Steps A 1 a: coloured image is decomposed into R passage, G passage and channel B;
Steps A 2: described R passage, G passage and channel B are carried out three layers of wavelet decomposition respectively and obtains corresponding wavelet conversion coefficient, wherein the low frequency of each layer of R passage and high fdrequency component distribute as shown in Figure 2, and other passages are also similar; Described R channel decomposition is R1 part, R2 part, R3 part and R4 part; Described G channel decomposition is G1 part, G2 part, G3 part and G4 part; Described channel B is decomposed into B1 part, B2 part, B3 part and B4 part;
Steps A 3: according to the openness distribution sample rate of described wavelet conversion coefficient, carry out described R passage, the block sampling of G passage and channel B, the detailed process of distributing sample rate is as follows: suppose that described coloured image is N × N coloured image, its sample rate is ratio, in R passage, R1 part accounts for 1/64 of total pixel, make the sample rate ratio1=0.9 of R1 part, in R passage, R2 part accounts for 3/64 of total pixel, make the sample rate ratio2=0.8 of R2 part, in R passage, R3 part accounts for 3/16 of total pixel, R4 part accounts for 3/4 of total pixel, the sample rate making R3 part is 2 times of R4 fractional-sample rate, i.e. ratio3=2 × ratio4, constant for ensureing the whole-sample rate of image, meet following formula:
Can be obtained fom the above equation:
,
The sample rate method of salary distribution as R passage is performed to described G passage and channel B.
Decode procedure comprises the following steps:
Step B1: described R1 part, G1 part, B1 part are merged into group1, described R2 part, G2 part, B2 part are merged into group2, described R3 part, G3 part, B3 part are merged into group3, described R4 part, G4 part, B4 part are merged into group4;
Step B2: sparse reconstruct is organized respectively to described group1, group2, group3 and group4;
Organizing sparse principle is give identical numbering to the same position of three passages, utilizes the correlation between them, finds the identical position of numbering in solving, and makes them have certain contact in this position.To be of a size of 16 × 16 coloured images, if a point block size is 2 × 2, then image can be divided into 64 blocks.Each image block is scanned the column vector of be 4 × 1, then can draw the definition procedure of group1 as shown in Figure 3.
Please refer to Fig. 3, (front 1/64 part is R1 to the wavelet coefficient first row first extracting in three passages respectively, G1, B1), stack with the form of column vector again, form 12 × 1 matrixes from R11 to B14, wherein R11 in Fig. 3, R12, R13, R14 is 4 elements of R1, G11, G12, G13, G14 is 4 elements of G1, B11, B12, B13, B14 is 4 elements of B1, group1 becomes identical numeral coordinate definition identical for relative position in 12 × 1 matrixes built up by three passages, such as R11, G11, B11 is defined as 1, R12, G12, B12 is defined as 2.When organizing sparse reconstruct, to reconstruct first pixel, need to find group1 matrix intermediate value be 1 coordinate (1; 5; 9), namely the reconstructed image meta of R, G, B triple channel composition is set to (1; 5; 9) pixel has correlation, and their value can be made similar, the principle of the sparse reconstruct of group that Here it is.In like manner, the definition of group2, group3 and group4 of 16 × 16 coloured images is respectively as shown in Fig. 4, Fig. 5, Fig. 6.
Then, SL20 algorithm is utilized to reconstruct the data of group1, group2, group3 and group4 respectively.Restructuring procedure and solving-optimizing problem
, wherein A is sensing matrix, and y is measured value, and x is wavelet coefficient,
for the wavelet coefficient of reconstruct, restructing algorithm flow process as shown in Figure 7.
Step B3: from carrying out four parts isolating three passages group1, group2, group3 and group4 after organizing sparse reconstruct, and resequence, obtains the wavelet coefficient of R passage, G passage and the channel B reconstructed; Wherein group1 isolates R1 part, G1 part and B1 part, wherein group2 isolates R2 part, G2 part and B2 part, wherein group3 isolates R3 part, G3 part and B3 part, and wherein group4 isolates R4 part, G4 part and B4 part (can with reference to figure 1);
Step B4: the picture content that wavelet inverse transformation obtains reconstruct is carried out respectively to described R passage, G passage and channel B;
Step B5: the combination of the picture content of described reconstruct is recovered described coloured image.
In order to verify the effect of the inventive method, for coloured image Peppers and Lena of 512 × 512, suppose that point block size is 32 × 32, calculation matrix is Bei Nuli matrix, color image compression of the present invention sampling and restructing algorithm and traditional RGB independence reconstructing method are compared, using Y-PSNR PSNR(unit for dB) as the reconstruct evaluation index of algorithm.Table 1 gives the PSNR of color image compression reconstructing method under low sampling rate.
Table 1:
As can be seen from Table 1, for coloured image Peppers and Lena, sample to image when equal sample rate, the method PSNR that the present invention proposes is than RGB independence reconstructing method height about 1-2dB.It can thus be appreciated that the compression sampling of the coloured image of the present invention's proposition and restructing algorithm when identical compression ratio, can effectively improve measurement efficiency and the reconstruction quality of coloured image.
The foregoing is only preferred embodiment of the present invention, all equalizations done according to the present patent application the scope of the claims change and modify, and all should belong to covering scope of the present invention.
Claims (2)
1. the compression sampling of coloured image and a restructing algorithm, is characterized in that:
Programming process comprises the steps:
Steps A 1 a: coloured image is decomposed into R passage, G passage and channel B;
Steps A 2: described R passage, G passage and channel B are carried out three layers of wavelet decomposition respectively and obtains corresponding wavelet conversion coefficient, wherein said R channel decomposition is R1 part, R2 part, R3 part and R4 part; Described G channel decomposition is G1 part, G2 part, G3 part and G4 part; Described channel B is decomposed into B1 part, B2 part, B3 part and B4 part;
Steps A 3: according to the openness distribution sample rate of described wavelet conversion coefficient, carries out the block sampling of described R passage, G passage and channel B;
Decode procedure comprises the following steps:
Step B1: described R1 part, G1 part, B1 part are merged into group1, described R2 part, G2 part, B2 part are merged into group2, described R3 part, G3 part, B3 part are merged into group3, described R4 part, G4 part, B4 part are merged into group4;
Step B2: sparse reconstruct is organized respectively to described group1, group2, group3 and group4;
Step B3: from carrying out four parts isolating three passages group1, group2, group3 and group4 after organizing sparse reconstruct, and resequence, obtains the wavelet coefficient of R passage, G passage and the channel B reconstructed;
Step B4: the picture content that wavelet inverse transformation obtains reconstruct is carried out respectively to described R passage, G passage and channel B;
Step B5: the combination of the picture content of described reconstruct is recovered described coloured image.
2. the compression sampling of coloured image according to claim 1 and restructing algorithm, it is characterized in that: being allocated as follows of sample rate in described steps A 3: suppose that described coloured image is N × N coloured image, its sample rate is ratio, in R passage, R1 part accounts for 1/64 of total pixel, make the sample rate ratio1=0.9 of R1 part, in R passage, R2 part accounts for 3/64 of total pixel, make the sample rate ratio2=0.8 of R2 part, in R passage, R3 part accounts for 3/16 of total pixel, R4 part accounts for 3/4 of total pixel, the sample rate making R3 part is 2 times of R4 fractional-sample rate, i.e. ratio3=2 × ratio4, constant for ensureing the whole-sample rate of image, meet following formula:
Can be obtained fom the above equation:
,
The sample rate method of salary distribution as R passage is performed to described G passage and channel B.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510442038.9A CN105072446B (en) | 2015-07-27 | 2015-07-27 | The compression sampling and restructing algorithm of a kind of coloured image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510442038.9A CN105072446B (en) | 2015-07-27 | 2015-07-27 | The compression sampling and restructing algorithm of a kind of coloured image |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105072446A true CN105072446A (en) | 2015-11-18 |
| CN105072446B CN105072446B (en) | 2018-05-04 |
Family
ID=54501712
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510442038.9A Active CN105072446B (en) | 2015-07-27 | 2015-07-27 | The compression sampling and restructing algorithm of a kind of coloured image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105072446B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106485760A (en) * | 2016-09-30 | 2017-03-08 | 电子科技大学 | A kind of coloured image Downsapling method based on minimum interpolation error quadratic sum |
| CN109997361A (en) * | 2016-12-21 | 2019-07-09 | 高通股份有限公司 | Low complex degree sign prediction for video coding |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090129466A1 (en) * | 2007-11-19 | 2009-05-21 | Samsung Electronics Co., Ltd. | Method, medium, and apparatus efficiently encoding and decoding moving image using image resolution adjustment |
| CN103346798A (en) * | 2013-06-05 | 2013-10-09 | 中国科学院微电子研究所 | Signal collecting method with sampling frequency lower than Nyquist frequency |
| CN103841583A (en) * | 2014-01-16 | 2014-06-04 | 华南理工大学 | Wireless network optimized mass signaling data collecting method based on compressed sensing |
-
2015
- 2015-07-27 CN CN201510442038.9A patent/CN105072446B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090129466A1 (en) * | 2007-11-19 | 2009-05-21 | Samsung Electronics Co., Ltd. | Method, medium, and apparatus efficiently encoding and decoding moving image using image resolution adjustment |
| CN103346798A (en) * | 2013-06-05 | 2013-10-09 | 中国科学院微电子研究所 | Signal collecting method with sampling frequency lower than Nyquist frequency |
| CN103841583A (en) * | 2014-01-16 | 2014-06-04 | 华南理工大学 | Wireless network optimized mass signaling data collecting method based on compressed sensing |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106485760A (en) * | 2016-09-30 | 2017-03-08 | 电子科技大学 | A kind of coloured image Downsapling method based on minimum interpolation error quadratic sum |
| CN106485760B (en) * | 2016-09-30 | 2019-05-14 | 电子科技大学 | A kind of color image Downsapling method based on minimum interpolation error quadratic sum |
| CN109997361A (en) * | 2016-12-21 | 2019-07-09 | 高通股份有限公司 | Low complex degree sign prediction for video coding |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105072446B (en) | 2018-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | An efficient deep convolutional neural networks model for compressed image deblocking | |
| Lin | High capacity reversible data hiding scheme based upon discrete cosine transformation | |
| CN102164282B (en) | Coefficient-random-permutation-based compressive sensing method and system for image coding | |
| CN101742306A (en) | High dynamic range codecs | |
| CN104199627B (en) | Gradable video encoding system based on multiple dimensioned online dictionary learning | |
| CN102708576A (en) | Method for reconstructing partitioned images by compressive sensing on the basis of structural dictionaries | |
| CN101971633A (en) | A video coding system with reference frame compression | |
| CN104065976B (en) | A kind of compression of images and security transmission method based on video | |
| CN111179144B (en) | Efficient information hiding method for multi-embedding of multi-system secret information | |
| CN107645662A (en) | A kind of colour image compression method | |
| CN103354617B (en) | Boundary strength compressing image quality objective evaluation method based on DCT domain | |
| CN106791836A (en) | It is a kind of to be based on a pair of methods of the reduction compression of images effect of Multi net voting | |
| Alam et al. | An improved JPEG image compression algorithm by modifying luminance quantization table | |
| CN105072446A (en) | Color image compression sampling and reconstruction algorithm | |
| CN112188217A (en) | JPEG compressed image decompression effect removing method combining DCT domain and pixel domain learning | |
| CN105163130B (en) | A kind of Lossless Image Compression Algorithm method based on discrete Tchebichef orthogonal polynomial | |
| Zhu et al. | Generative high-capacity image hiding based on residual CNN in wavelet domain | |
| CN105072453B (en) | A kind of video watermark process of facing moving terminal | |
| CN102572426B (en) | Method and apparatus for data processing | |
| CN107920250B (en) | Compressed sensing image coding transmission method | |
| CN104683818B (en) | Method for compressing image based on biorthogonal invariant set m ultiwavelet | |
| CN103763566A (en) | Color halftone image compressing method based on three-dimensional matrix WDCT transformation | |
| Sadkhan | A proposed image compression technique based on DWT and predictive techniques | |
| CN101257629A (en) | Encoding and decoding method of image sequence nondestructive compression based on interested area | |
| CN103413336A (en) | Grid non-aligned double-JPEG-compression detecting method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |