In image compressive sensing

we have to take it all into the in-depth study for solving the problem of reconstruction effect and time-consuming. Based on discrete cosine sparse basis

we choose random gaussian matrix as observed sampling

In view of the better reconstructed image but Basis Pursuit reconstruction algorithm perform slowly. combining with the merits of the image block can improve the precision of image

we put forward a kind of Basis Pursuit reconstruction algorithm based on image block

then the traditional OMP algorithm、Basis Pursuit algorithm、Compressive sampling algorithm、Compressive sampling algorithm based on image block

OMP algorithm and BP algorithm based on Overcomplete Dictionaries for Sparse Representation are compared. Under different sampling rate we perform MATLAB simulation experiment

and we get the reconstructed image with the peak signal to noise ratio of the reconstructed image and the consuming time. The experiment results show that not only the Basis Pursuit algorithm based on image block is 1 to 10 d B higher at the peak signal-to-noise ratio

but also is shorter at the running time than others. Thus the Basis Pursuit algorithm based on image block is the best one

what's more

considering the question that how to set the value of image block is a good choice

lots of same experiments are done

getting the solution that every block is setted as 8*8 and sparsity that is equal to(0.15～0.4) of the old dimensions are the better choices.