A High Performance Image Coding Using Uniform Morphological Sampling, Residues Classifying, and Vector Quantization

In this paper a new image coding scheme based on the uniform morphological sampling is presented. In the proposed algorithm, the image is sub-sampled uniformly using a sampling grid of squares of size 4 in Heijmans method. The sampling process is equivalent to decomposing the image into 4×4 blocks and each block is represented by its minimum intensity (sample value). The residual blocks are then classified into uniform and non-uniform blocks according to a discrete gradient. The uniform blocks are represented by their mean value. Each non-uniform block is represented by its minimum value and a block (vector) chosen among a predetermined codebook blocks (vectors). The uniform and non-uniform blocks are coded by a different number of bits. Also, a hierarchical version is proposed which provides a higher compression ratio for an approximately equivalent visual quality. Several experiments are made, and compression ratios of 22.64 to 25.19 for a good visual quality of reconstructed images are obtained.