A window-based color quantization technique and its architecture implementation

A new color quantization (CQ) technique and its VLSI implementation is introduced. It is based on splitting an image into windows and uses a Kohonen self organized neural network classifier (SONNC). Initially, the dominant colors of each window are extracted through the SONNC and then are used for the quantization of the colors of the entire image. The image split into windows offers reduction of the memory requirements and feasibility of a suitable VLSI implementation of the most time consuming part of the technique, i.e. the SONNC. Applying a systematic design methodology into the developed CQ algorithm, an efficient system-on-chip based on an ARM (Advanced RISC Machines Ltd) processor, which achieves high speed processing and less energy consumption, is derived.