PROJECT TITLE :

Approximation and Compression With Sparse Orthonormal Transforms - 2015

ABSTRACT:

We tend to propose a brand new transform style method that targets the generation of compression-optimized transforms for next-generation multimedia applications. The fundamental plan behind rework compression is to exploit regularity inside signals such that redundancy is minimized subject to a fidelity cost. Multimedia signals, in specific pictures and video, are well-known to contain a various set of localized structures, leading to many different sorts of regularity and to nonstationary signal statistics. The proposed method designs sparse orthonormal transforms (SOTs) that automatically exploit regularity over completely different signal structures and provides an adaptation technique that determines the most effective representation over localized regions. Unlike earlier work that's motivated by linear approximation constructs and model-based styles that are limited to specific types of signal regularity, our work uses general nonlinear approximation ideas and a knowledge-driven setup to significantly broaden its reach. We show that our SOT styles give a safe and principled extension of the Karhunen-Loeve transform (KLT) by reducing to the KLT on Gaussian processes and by automatically exploiting non-Gaussian statistics to considerably improve over the KLT on additional general processes. We give an algebraic optimization framework that generates optimized designs for any desired remodel structure (multiresolution, block, lapped, and so on) with significantly higher n-term approximation performance. For every structure, we have a tendency to propose a brand new prototype codec and take a look at over a database of images. Simulation results show consistent increase in compression and approximation performance compared with conventional methods.