Inhomogeneous Measurement High-speed Reconstruction Method of Image Block Compressed Sensing

ABSTRACT

The complexity of block compressed sensing of images algorithm is very high. And it is too difficult to apply the algorithm on hardware platform. A low complexity compressed sensing method based on non-uniform measurement is presented in this paper which use different measurement to the transform domain coefficient with different significance of every image blocks. And as the experimental results on PC platform suggest, the proposed method is better than original method on reconstructed image’s subjective and objective quality. And this method is optimized and implemented on a high performance digital signal processor TMS320C6455 platform including algorithm structure design and C code optimization according to the hardware feature. It is also faster than original method and it get about 4 times speedup on the DSP platform after optimization. The proposed method can be combined with the existing image and video compression scheme very easily.

Keywords:

Image compressed sensing, digital signal processor, optimized implementation, random measurement, image reconstruction.