An Analysis of Block Sizes for Compressive Sensing Reconstruction Applied in Image Processing Optimisation

Rubens M. B. da S. Lima, Hugo B. S. Araujo, Cleonilson P. de Souza

Abstract

Signal sampling is a fundamental process of data acquisition systems and several studies have emerged regarding sampling methods following on Nyquist Theorem. Compressive Sensing (CS) proposes sampling of sparse signals with sub-Nyquist sampling rates. In short, CS is composed of a sampling/compress stage and a reconstruction stage. Some algorithms are used where this last. One of these is the Orthogonal Matching Pursuit (OMP) where 2D Discrete Cosine Transform (2D-DCT) can be used. This work compared the application of 2D-DCT transform and CS theory on images as either a whole or split in blocks. As a result, the influence of block size is revealed using the Mean Square Error (MSE) metric for different block sizes.

Keywords

Compressive Sampling; Image Compression; DCT; Block Size Influence; Innovation

Download

IMEKO-TC10-2022-012.pdf

DOI

10.21014/tc10-2022.012

IMEKO TC

TC10 - Measurement for Diagnostics, Optimization and Control

Back to the proceedings