Pyramidal Image Compression Using the Parameterized NEDI Algorithm
Mikhail V. Gashnikov · Optical Memory and Neural Networks · 2021
Abstract In the scope of pyramid image compression we offer the parameterization of NEDI-based interpolation algorithm. The original NEDI is an adaptive algorithm because it calculates the weight coefficients at each image point. The algorithm selects parameters that are best for the interpolation within the given vicinity of a particular pixel. The article proposes the parameterization of the NEDI algorithm which allows higher efficiency of the algorithm due to its better adaptiveness. Based on the evaluation of the intensity and direction of a local irregularity at each pixel, the parameterization makes it possible to simplify the structure of the interpolation algorithm and lessens its intricacy. The parameterized NEDI algorithm we offer is considered as a part of the pyramidal image compression method, which uses slightly thinned levels of the pyramidal representation to interpolate heavily thinned levels of the same pyramidal representation. Computational experiments prove that the use of the parameterized NEDI algorithm improves the efficiency of the pyramidal image compression method.