The Research On Compressive Sensing Of Color Image

In the age of rapid development of information technology,people's demands for the information are also growing.The main way we get information is through vision.It means image information is very importantto us.Color images provide richer information than grayscale images,it is convenient for people to receive.But the data of color image is greater than grayscale images,so the color image compression,storage and transmission are important issues need to be addressed.Many research scholars have been trying to explore more effective ways to improve the compression ratio in the premise of ensuring the image quality.Compressed sensing(CS)theory emerged in recent year states that it is with a high probability to reconstruct the original unknown sparse signal accurately from a small number of measurements.It breaks through the limitation of Shannon sampling theorem,signal sampling frequency not allowed to be less than twice the highest frequency of the analog signal.Therefore compressed sensing theory brings new ideas to the color image compression.This paper focuses on the study of color image compressed sensing,the main work is as follows:(1)Describes basic knowledge of image as well as the characteristics of color image,introduces the wavelet transform theory and multi-scale decomposition of wavelet,expound the principle of compressed sensing and perception of color image compression method.(2)Aiming the problem how to select appropriate sampling numbers for a color image signal with unknown sparsity level.According to particular fingerprint information of color image,we establish a new stopping rule and design a method named sequential compressed sensing method based on the fingerprint of color image.In this method,a color image is decomposed into separate RGB components in the perception end and recovered in the reconstruction end.The stopping rule controls the interactive process between the two ends,to achieve the purpose for different signal adaptively select the appropriate sampling numbers.Experimental results show that,our proposed method can improve the quality of reconstructed color image.(3)Noting that the color channels are highly correlated and that the high-frequency wavelet coefficients in three directions at the same level after the multi-scale wavelet transform are also highly correlated,we propose a new method named color image compressed sensing based on Bayesian multi-task.In the signal reconstruction phase,reconstruct the multiple signals with high correlation jointly using Bayesian multi-task.Experimental results show that our proposed algorithm based on correlation multi-tasking can improve quality of the color reconstructed image.