The Research Based On The Zero-tree Wavelet Image Compressive Sensing Method

Mankind has stepped in the era of information technology in an all-round way today,more and more researchers pay much attention to image information field,especially in military,aerospace,medical care etc.In recent years,Candes and Donoho has proposed the compressed sensing(CS)theory,which makes good use of the compressibility and sparstity of the signal,then CS makes initial signal get its low dimensional projection,finally,we get the reconstruction of initial signal by the appropriate signal recovery algorithm.CS breaks through the limitation of sampling frequency on traditional signal sampling theorem,achieves much lower than the Nyquist sampling frequency of the signal,It makes signal sampling and compression perform at the same time,which improves the transmission efficiency and provides a new way for the traditional signal processing methods.This paper mainly combines structure of image wavelet coefficients with zero-tree coding,and makes further research on image compressed sensing.The main work is as follows:(1)We propose a method of image compressed sensing based on wavelet zerotree coding.After the analysis and research on the correlation of father and son coefficient of image wavelet transform,we integrate the wavelet zero-tree coding into the measurement and reconstruction steps in compressed sensing.In the measurement step,we design a kind of two symbol zero tree coding algorithm based on threshold for the image wavelet coefficient,and the sparse signal is coded and compressed sampling at the same time.In the reconstruction step,we recover the position index of large coefficient in the way of zero-tree decoding,then propose an kind of zero-tree pursuit algorithm to get the reconstruction.Experimental results show that,compared to the traditional CS,image compressed sensing based on wavelet zero-tree that our proposed doesn't only improve the efficiency of reconstruction,transmission and the compression ratio,but also the processing time is shorten.(2)Combining the wavelet coefficient quantization steps with sequential compressed sensing principle,we propose a kind of embedded image compression sensing method.In this method,we send sampling signal and the string quantified from the transmission end to the reconstruction end,establish a stopping andfeedback rule to ensure the appropriate interaction between the transmission end to the reconstruction end.Our proposed method doesn't only achieve to sampling on the unknown original signal as sequential CS,but also reduced the size of the sample matrix and the storage cost.The experimental results show that,compared to the traditional CS and sequential CS,our proposed method show obvious superiority in the the processing time adaptive adjustment of the growth of the measurement.