Multiresolutional piecewise-linear image decompositions: quantization error propagation and design of "stable" compression schemes

Abstract

Summary form only given. The paper introduces a new approach to design of stable tile-effect-free multiresolutional image compression schemes. It focuses on how quantization errors in the decomposition coefficients affect the quality of the decompressed picture, how the errors propagate in a multiresolutional decomposition, and how to design a compression scheme where the effect of quantization errors is minimized (visually and quantitatively). It also introduces and analyzes the simplest family of Laplacian pyramids (using 3-point causal filters) which yield multiresolutional piecewise-linear image decompositions. This gives reconstructed images much better visual appearance without blockiness, as the examples. The error propagation analysis has lead to discovery of particular Laplacian pyramids where quantizations errors do not amplify as they propagate, but quickly decay.