New quantization matrix - embedded in discrete cosine transform, for real-time encoding/decoding system (REDS)

Abstract

A new quantization matrix is proposed for large sized HDTV image. By the rate and distortion tradeoff, an optimal quantization matrix is designed, and to implement in Real time Encoding/Decoding System (REDS), the scaleable quantization matrix has to be used. Because image dependent quantization matrices are not suitable for real time processing, a new quantization matrix that can be used image-independently, is proposed. And the performance of the proposed quantization matrix is simulated with 31 HDTV images with size of 1920×1080. The simulation results represent the visible blockiness of the reconstructed images with proposed quantization matrix is less than those with conventional JPEG default quantization matrix. Also, the PSNR to entropy graphs represent the proposed quantization matrix achieves better quality than the conventional matrix. Also the quantization matrix is embedded in Discrete Cosine Transform (DCT) to implement in REDS. Therefore, there are no additional cycles to perform the quantization, and it makes REDS can be perform the encoding/decoding processing in real-time.