Decomposição triangular de imagens: uma aplicação em compressão.

Abstract

The objective of this w o r k is to bring a contribution to the field of image compression, i . e., the problem of reducing the amount of bits needed to represent a given imge, w h i l e still being able to restore i t to its shape (to w i t h i n a certain degree of fidelity), whenever needed. A good solution to i t is important because i t can represent a considerable economy i n terms of image storage space and/or transmission bandwidth. This w o r k presents a new method for image compression, using Tritree (TT) decompositon. TT decomposition is similar to Quadtree (QT) decomposition, which has been largely used for image compression algorithms. However, w h i l e QT works by subdividing the image into progressively smaller square regions, TT decompositon does i t for triangular regions. The goal is to segment the image into a set of triangular homogeneous regions, where the pixel values difference does not exceed a given threshold. A tree is b u i l t to represent the decomposition. Each triangle w i l l be a node of the TT The i n i t i a l triangle, called root, containing the whole image. The final triangles, representing the compressed image, are the leaves. Experimental results show the proposed method to perform better than the QT decomposition, i n terms of reconstructed images and compression rate.