Image compression methods with distortion controlled capabilities
Traditionally, lossy compression schemes have focused on compressing data at fixed bit rates in order to communicate information over limited bandwidth communication channels, or to store information in a fixed-size storage media. In this paper we present a class of lossy data compression algorithms that are capable of encoding images so that the loss of information complies with certain distortion requirements. The developed algorithms are based on the Tree-Structured Vector Quantizers (TSVQ). The first distortion controlled algorithm uses variable-size image blocks, encoded on quad-tree data structures, to efficiently encode image areas with different information content. Another class of distortion controlled algorithms that are presented is based on recursive quantization of error image blocks that represent the difference between the current approximation and the original block. We will also describe the progressive compression properties of these algorithms. Finally, we will present their compression/distortion performance using satellite images provided by NASA, and we will show that they achieve better performance than the TSVQ algorithms at high bit rates.
