Joint rate-distortion optimization of discrete cosine transform coefficient quantization and disparity field estimation for stereo image coding

Abstract

Disparity-compensated residual coding (DCRC) is the most popular approach to stereo image coding because of its simplicity and good performance. Researches on DCRC have investigated two main topics: quantization of discrete cosine transform (DCT) coefficients and disparity field estimation and compensation. It is shown that quantization problems can be optimized in a rate-distortion (RD) sense under the fixed-disparity constraint and that the RD-optimized disparity field can be obtained under the fixed-quantizer constraint. This paper proposes a technique that jointly optimizes DCT coefficient quantization and dispar- ity field estimation using the three-dimensional Viterbi algorithm with re- spect to the RD theory under three constraints: epipolar constraint, inter- block disparity constraint, and intrablock disparity constraint. The proposed algorithm reduces not only the bit rate of the disparity map but also that of DCT coefficients, leading to performance improvement. Also, it overcomes shortcomings of conventional methods and is well applied to the narrow-angle stereo camera systems. The effectiveness of the proposed algorithm is shown via computer simulations. © 2002 Society of