Motion estimation and compensation is widely used for exploiting temporal correlation within an image sequence. To find motion vectors that lead to high compression, most motion estimation approaches use a source distortion measure, such as mean-square error (MSE) or mean-absolute error (MAE), as a search criterion. When incorporated into a closed-loop motion compensated (MC) transform video coder, these schemes produce noisy motion fields which significantly increase the bit-rates required to represent motion vectors. In view of this problem, this paper presents a rate-distortion optimal motion estimation algorithm. The proposed scheme improves rate performance of the estimated motion field while maintaining the peak signal-to-noise ratio (PSNR) prediction quality of the distortion-based methods, thereby enabling an efficient bit allocation between motion information and transform-coded prediction residuals. For coders in which motion vectors are differentially encoded, the rate-distortion optimization process is formulated as a shortest-path-finding problem. Adopting this framework, we show that the optimal solution for the conventional block-based motion estimation, followed by one-dimensional (1-D) differential coding and Huffman coding, can be obtained by using dynamic programming or the Viterbi algorithm. We propose an effective fast algorithm that closely approximates the optimal performance while requiring considerably less complexity. Our experimental results demonstrate overall gains in the range of 0.3-1.5 dB.

Rate-distortion optimal motion estimation algorithms for motion-compensated transform video coding

We describe the RD-OPT algorithm for DCT quantization optimization, which can be used as an efficient tool for near-optimal rate control in DCT-based compression techniques, such as JPEG and MPEG. RD-OPT measures DCT coefficient statistics for the given image data to construct rate/distortion-specific quantization tables with nearly optimal tradeoffs.

An efficient algorithm for optimizing DCT quantization

We describe a trellis-based algorithm which enables R-D optimum quantization decisions in the H.263+ video coding standard. The algorithm allows the quantization decisions for all coefficients in a block to be made jointly, and experiments conducted using H.263+ for video coding rates of 40-50 kbps show an average improvement of 3.5% in bitrate relative to implementations which follow the International Telecommunications Union (ITU) Test Model specifications in making quantization decisions.

Trellis-based R-D optimal quantization in H.263+

We describe a trellis-based algorithm which enables R-D optimum quantization decisions in the H.263+ video coding standard. The use of the trellis allows the quantization decisions for all coefficients in a block to be made jointly, and contrasts with more commonly used R-D optimizations which operate on one coefficient at a time. Experiments conducted using H.263+ for video coding rates of 40-50 kbps show an average improvement of 3.5% in bit rate, or equivalently 0.17 dB in PSNR, relative to implementations which follow the International Telecommunications Union (ITU) Test Model specifications in making quantization decisions.

The quantization process is a main part of image compression to control visual quality and the bit rate of the image output. The JPEG quantization tables are obtained from a series of psychovisual experiments to determine a visual threshold. The visual threshold is useful in handling the intensity level of the colour image that can be perceived visually by the human visual system. This paper will investigate a psychovisual error threshold at DCT frequency on the grayscale image. The DCT coefficients are incremented one by one for each frequency order. Whereby, the contribution of DCT coefficients to the error reconstruction will be a primitive pyschovisual error. At certain threshold being set on this psychovisual error, the new quantization table can be generated. The experimental results show that the new quantization table from psychovisual error threshold for DCT basis functions gives better quality image at lower average bit length of Huffman code than standard JPEG image compression.

A generic psychovisual error threshold for the quantization table generation on JPEG image compression

Previous works, including adaptive quantizer selection and adaptive coefficient thresholding, have addressed the optimization of a baseline-decodable JPEG coder in a rate-distortion (R-D) sense. In this work, by developing an entropy-constrained quantization framework, we show that these previous works do not fully realize the attainable coding gain, and then formulate a computationally efficient way that attempts to fully realize this gain for baseline-JPEG-decodable systems. Interestingly, we find that the gains obtained using the previous algorithms are almost additive. The framework involves viewing a scalar-quantized system with fixed quantizers as a special type of vector quantizer (VQ), and then to use techniques akin to entropy-constrained vector quantization (ECVQ) to optimize the system. In the JPEG case, a computationally efficient algorithm can be derived, without training, by jointly performing coefficient thresholding, quantizer selection, and Huffman table customization, all compatible with the baseline JPEG syntax. Our algorithm achieves significant R-D improvement over standard JPEG (about 2 dB for typical images) with performance comparable to that of more complex "state-of-the-art" coders. For example, for the Lenna image coded at 1.0 bits per pixel, our JPEG-compatible coder achieves a PSNR of 39.6 dB, which even slightly exceeds the published performance of Shapiro's wavelet coder. Although PSNR does not guarantee subjective performance, our algorithm can be applied with a flexible range of visually-based distortion metrics.

JPEG optimization using an entropy-constrained quantization framework

We introduce a novel image-adaptive encoding scheme for the baseline JPEG standard that maximizes the decoded image quality without compromising compatibility with current JPEG decoders. Our algorithm jointly optimizes quantizer selection, coefficient 'thresholding', and entropy coding within a rate-distortion (R-D) framework. It unifies two previous approaches to image-adaptive JPEG encoding: R-D optimized quantizer selection by Wu and Gersho, and R-D optimal coefficient thresholding by Ramchandran and Vetterli. By formulating an algorithm which optimizes these two operations jointly, we have obtained performance that is the best in the reported literature for JPEG-compatible coding. In fact the performance of this JPEG coder is comparable to that of more complex 'state-of-the-art' image coding schemes: e.g., for the benchmark 512 by 512 'Lenna' image at a coding rate of 1 bit per pixel, our algorithm achieves a peak signal to noise ratio of 39.6 dB, which represents a gain of 1.7 dB over JPEG using the example Q- matrix with a customized Huffman entropy coder, and even slightly exceeds the published performance of Shapiro's celebrated embedded zerotree wavelet coding scheme. Furthermore, with the choice of appropriate visually-based error metrics, noticeable subjective improvement has been achieved as well. The reason for our algorithm's superior performance can be attributed to its conceptual equivalence to the application of entropy-constrained vector quantization design principles to a JPEG-compatible framework. Furthermore, our algorithm may be applied to other systems that use run-length encoding, including intra- frame MPEG and subband or wavelet coding.© (1996) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Matthew S. Crouse

Papers

JPEG optimization using an entropy-constrained quantization framework

Joint thresholding and quantizer selection for block-transform-based image coding with applications to baseline JPEG

JPEG Optimization Entropy-Constrained Quantization Framework.