Implementation and Performance Assessment of Compressed Sensing for Images and Video Signals

Abstract

Compressed sensing is an optimization based formalized framework based upon sub-Nyquist sampling principle of exploiting only the sparse signal of interest. It exploits the sparsity of the signal to reconstruct it from less number of measurements than required by the Nyquist sampling criteria. A nascent field of compressive sensing is explored in this paper for accurate acquisition and reconstruction of signals, images and video sequences.  The algorithm is proposed for compression and efficient recovery of image and video based on the concept of compressive sensing. Three basic reconstruction techniques (Basic Pursuit (l₁) Minimization, Least Square (l₂) Minimization and Orthogonal Matching Pursuit) are applied on image samples and they are compared based on quality performance criteria. The performance parameters like compression ratio, peak signal to noise ratio and structural similarity index are evaluated for different image and video samples for critical analysis of these performance parameters is done for different reconstruction schemes. Finally it is concluded that compressive sensing based approach is better than the traditional compression schemes and Basic Pursuit (l₁) methods gives the better image quality with a tradeoff among other parameters enabling faster acquisition, compression and reconstruction.

Keywords: Compressed sensing, Sparsity, Structural similarity index, Basic Pursuit Minimization, Least Square Minimization and Orthogonal Matching Pursuit.