Adaptively Encoding Of Compressive Sensing For Image Signal

Recently, an emerging sampling theory of signal acquirementnamed Compressive Sensing(CS) proposed by Candes and Donohoprovides important theoretical basis for signal compression sampling.Different from the conventional Nyquist sampling, the signals aresampled at a rate at least twice the highest frequency present in thesignals. Compressive sensing theorem indicates that, if the signals aresparse and compressible, signals can be recovered without distortion oncondition that sampling signal ration is far lower than samplingfrequency based on Nyquist. Compressive Sensing has shown goodapplication prospect in signal acquistion, communication, imageprocessing, computer vision and other fields. Image compression basedon compressive sensing has also become the research hotspot of thecurrent image coding.This thesis focus on the exploratory study of image compressionbased on compressive sensing, and study the encoding optimizationmethod based on image compressive sensing to effectively improve themeasuring efficiency and the quality of image reconstruction based onimage representation of compressive sensing. The main works andcontributions of this thesis are as follows:(1) Reweighted compressive sensing encoding methods in blockDCT domain are investigated thoroughly. Considering that the structurecharacteristics of the image signals are not considered into the measurematrixs of conventional compressive sensing, it introduces a reweightingscheme into the conventional CS framework whose coefficients aredetermined in encoding side according to the statistics of image signals.Experimental results demonstrate that this method notably improvemeasuring efficiency of the system.(2) Combined with the human perception characteristics and considering that human eyes are more sensitive to low frequencycomponents than to high frequency ones, A novel adaptivelycompressive sensing encoding for image signal is proposed based onsubband energy distribution characteristics. The proposed adaptivelyreweighted compressive sampling method could not only efficientlyreduce the computational complexity, but also considerably decreasemeasurement rate and enhance the recovery image quality in both PSNRand subjective visual quality.