Digital image coding techniques: Codec design using vector quantization method.

摘要

Recently there have been many efforts to improve coding systems for digital images and speech signals. Among the various techniques, a vector quantization (VQ) technique has proven that it can achieve a high compression ratio without sacrificing subjective and objective quality of an image.; During the past few years several design algorithms have been developed for a variety of vector quantizers and the performance of coders has been studied. The purpose of this research is to analyze some of these design techniques and facilitate their applications to digital image coding. The goal of this research project targeted and achieved a lower bit rate while maintaining image quality. Artificial neural networks such as Kohonen's self-organizing feature map (KSFM) and its variant have been applied and developed during this research. Several new encoding and decoding (codec) processes have been developed to maintain a high compression ratio (from 8 bits/pixel to 0.28 bits/pixel in test image Lena) without losing subjective and objective quality. I have obtained several experimental results which demonstrate the improvement of the new vector quantizers, making them comparable to industry standard and existing vector quantizers.; In recent years, the demand for digital video transmission and storage has increased dramatically in applications to medical images, teleconferencing (CCITT H.261), multi-media systems (MPEG-I, MPEG-II), and HDTV. In order to minimize the memory for storage and the bandwidth for transmission, digital video compression techniques with low implementation complexity have become crucial and mandatory. For digital video coding, a new vector quantization scheme (MCVQ) is proposed for interframe predictive coding of images to achieve a high compression ratio. Quantization method consists of a series of pipe-line processes. For frame sequences, motion compensation is used to reduce the variance of input vectors. Several codec simulations have been implemented and compared with industry standard such as MPEG and CCITT H.261. In this research a difference vector quantizer and subband coding (SBC) techniques (for the second layer of implementation) have been applied. The simulations show satisfactory results in implementation complexity and high compression ratio (from 60.7 Mbits/second to 1.2 Mbits/second) for video frame sequences.