Sensed compression with cosine and noiselet measurements for medical imaging

Abstract

We verified compression efficiency of the procedures based on compressive sensing (CS) inspiration. Medical imaging was concerned as challenging area of possible applications. Irreversible image compression algorithm was integrated with source data measurements according to CS rules. Two kinds of measurements as inner products against adjusted atoms were used: regular cosines and pseudo-random noiselets. Image coarse representation was approximated from linear cosine measurements while important details were estimated basing on fixed noiselet measurements. Simulated sensor system projected the image onto a set of separable 2-D basis functions to measure the corresponding expansion coefficients. Such procedure was optimized and augmented to construct integrated method of image sensing, compression and data processing. We proposed algorithm of selected measurements with uniformly quantized coefficients formed and encoded with necessary side information. Universal PAQ8 archiver was used to complete compression procedure. Experimentally verified compression schemes showed possible compression improvement by designed procedure in comparison to reference JPEG and JPEG2000 encoders. © 2012 Springer-Verlag.