Optimal Quantization and Adaptive Interpolation in Compression of Multidimensional Signals
Mikhail V. Gashnikov · Optical Memory and Neural Networks · 2020
Abstract The paper deals with the algorithms of optimal quantization and adaptive interpolation in interpolative and hierarchical compression of multidimensional signals. An uneven quantization scale optimization algorithm is invented to tackle the unknown number of quantization levels given soft requirements for the optimization criterion. An important instance of applying the quantizer to the compression problem is considered. Additionally, the adaptation of parameterized interpolation algorithms to interpolative and hierarchical compression methods is made that includes the definition of interpolation functions, inference rules for these functions, and inference rules optimization algorithms. The optimizers and quantizers are tested in the compression of real multidimensional data.