A study on source coding of super definition images with vector quantization

A vector quantization scheme specifically designed for super high definition images is proposed. This coding scheme is organized in two phases. In the first phase, multistage vector quantization takes place in the subband domain. In the second phase, threshold block selective vector quantization is done to reduce the visible blockiness distortion. The input image has a resolution of 2048 by 2048 pixels and is first divided into four bands and decimated using vertical and horizontal filtering. In each band, the decimated image signal is quantized using multistage vector quantization. In each vector quantization stage, either a self-learning address dictionary method or a threshold block selective method is utilized. High quality and definition decoded images are obtained through this method. Its quality is examined using a super high definition image display system.