Overlapped block-based compressive sensing imaging on mobile handset devices

Abstract

Compressive Sensing (CS) is a new technique that simultaneously senses and compresses an image by taking a set of random projections from the underlying scene. An optimization algorithm is then used to recover the initial image. In practice, these optimization algorithms have restricted CS techniques to be implemented on high performance computational architectures, such as personal computers or graphical processing units (GPU) due the huge number of operations required for the image recovery. This work extends the application of CS to be implemented in an extremely limited memory and processing architecture such as a mobile device. Specifically, overlapped blocking-based algorithms are developed such that it is possible to reconstruct an image on a mobile device. An analysis of the energy consumption of the block-based CS algorithms is presented. The results show the required computational time for reconstruction and the image reconstruction quality for images of 128x128 and 256x256 pixels.