Wavelet-Based Quality Constrained Compression Using Binary Search

Recently Gao et al. proposed a wavelet-based image quality metric, based on which, a quality constrained image subband quantization scheme is derived for lossy compression. But this scheme needs entropy coding to find the compression ratio decrement for tuning the final quantization step of each subband, which is time consuming and wasteful. In addition, computing the initial quantization steps also relies too much upon parameter tuning. This paper is a follow-up study, and proposes a simple binary search to determine the quantization step, according to the observation that, the quality metric decreases monotonically as the quantization steps increase. Experiments show that, the proposed quantization method outperforms its old counterpart in terms of computing expense, image quality, compression ratio and mathematical elegance. Keywords— Lossy compression, quantization steps, Image Quality Metric, binary search, inverse function