Research On Robust Methods Of Image Processing Based On Compressed Sensing

In the information age, the amount of information carried by the image is fargreater than the voice data, The image data’s practical, intuitive, specific a vivid, highefficiency, these characteristics determines the image communication will become oneof most important means of communication. The conventional digital imagecompression technology for efficient storage target method has been unreliablechannel suited to multi-service digital image transmission. To enhance thecompressed data appears lost and prevent data loss of the reconstructed image qualitydegradation phenomenon has also become a focus for researchers. CompressiveSensing (Compressed Sensing), a breakthrough in the two difficulties in the field oftraditional image signal processing: the Nyquist criterion of sampling frequency toohigh qualified, while improving the hardware requirements; signal processingframework sampled compression low efficiency, and wasting storage space. TheCompressive Sensing technology research focuses on the sparse representation of theimage, select the measurement matrix, as well as improved recovery algorithm forimage compression.In this paper, based on compressed sensing image processing, a block-basedcompression sampling mode compression for natural images sparse transformedperceptual processing program is presented. Connection with the corresponding modelCS recovery algorithm in image restoration capabilities at the same time CS canimprove the robustness of the image compression processing. The major contributionsare four-fold as follows:1) Wavelet transform DWT as the most commonly used image sparse transformhas been widely used in image compression algorithm which is also used in acompressed sensing deal with them as an image signal sparse representation. In thispaper, do experiments and analysis on wavelet coefficients energy distribution andcoefficient correlation. Coefficients which are low-frequency coefficients amplitudecontains most of the energy of the image, this part of the coefficient is very important,but not enough sparse sampling conditions. These coefficients are the key on imagerestoration and the robustness of image. Effective treatment of low-frequencycoefficients is the key to improve image compression perception of treatment effect.2) Research on image block technique of compressed sensing image processing. Multi-dimensional image data is compressed when the perception process needs to beconverted into one-dimensional data encoding measurement when measuring the sizeof the matrix depends on the length of the one-dimensional signal, when the imageprocessing in a CS measurements require very large matrix to be measured, themeasurement process will consume a lot of storage space to save the measurementmatrix, measurement speed plummeted recovery algorithm efficiency will be reduced,the original time complexity of a higher degree of recovery calculation will becomemore slowly, so the effective image block will improve image compression perceptionof the efficiency of the treatment and recovery effect, the existing block directly to theoriginal image block, which will produce the corresponding block effect and theintroduction of noise, compared to the corresponding sub-block method and proposeda new sub-block way, and the corresponding recovery algorithms, the feasibility ofthis approach is verified by experiments.3) Research on modeling compressed sensing recovery algorithm, the imageblock technology in conjunction with the corresponding recovery algorithms in orderto better image recovery effect, contrast block sparse model-based BOMP recoveryalgorithms and experimental.4) Research on compressed sensing image signal processing robustness(anti-jamming capability) enhancement. The original image after random sampling,each of information is equally important, even if some items which are lost, stillperfect reconstruction without adding in the coding flow correction, error detectioncode can still ensure robustness. Study and verify the robustness of imageenhancement image coding based on compressed sensing.