Análise de um Algoritmo para a Compressão de Imagem utilizando a técnica SVD Aleatório

Abstract

This work presents the implementation and evaluation of an image compression algorithm based on Randomized Singular Value Decomposition (Randomized SVD). The method was developed in Python, focusing on grayscale BMP images, aiming to reduce storage size without excessively compromising visual quality. Compression tests were performed with different control values (k), allowing analysis of the parameter’s influence on the compression ratio and bitrate. The results showed that, although significant compression was achieved, the visual quality was maintained at a reasonable level, sufficient for human understanding but with a noticeable loss of detail. Furthermore, it was observed that the method is not applicable to already compressed images (such as JPEG), color images (RGB), or videos. For future studies, it is recommended to adapt the algorithm for color images and videos, expanding its applicability to multimedia contexts.