Abstract: The notion of perceptual coding, which is based on the concept of distortion masking by the signal being compressed, is developed. Progress in this field as a result of advances in classical coding theory, modeling of human perception, and digital signal processing, is described. It is proposed that fundamental limits in the science can be expressed by the semiquantitative concepts of perceptual entropy and the perceptual distortion-rate function, and current compression technology is examined in that framework. Problems and future research directions are summarized. >

Abstract: Fractals are geometric or data structures which do not simplify under magnification. Fractal Image Compression is a technique which associates a fractal to an image. On the one hand, the fractal can be described in terms of a few succinct rules, while on the other, the fractal contains much or all of the image information. Since the rules are described with less bits of data than the image, compression results. Data compression with fractals is an approach to reach high compression ratios for large data streams related to images. The high compression ratios are attained at a cost of large amounts of computation. Both lossless and lossy modes are supported by the technique. The technique is stable in that small errors in codes lead to small errors in image data. Applications to the NASA mission are discussed.

Abstract: Many image compression standards (JPEG, MPEG, H.261) are based on the Discrete Cosine Transform (DCT). However, these standards do not specify the actual DCT quantization matrix. We have previously provided mathematical formulae to compute a perceptually lossless quantization matrix. Here I show how to compute a matrix that is optimized for a particular image. The method treats each DCT coefficient as an approximation to the local response of a visual 'channel'. For a given quantization matrix, the DCT quantization errors are adjusted by contrast sensitivity, light adaptation, and contrast masking, and are pooled non-linearly over the blocks of the image. This yields an 8x8 'perceptual error matrix'. A second non-linear pooling over the perceptual error matrix yields total perceptual error. With this model we may estimate the quantization matrix for a particular image that yields minimum bit rate for a given total perceptual error, or minimum perceptual error for a given bit rate. Custom matrices for a number of images show clear improvement over image-independent matrices. Custom matrices are compatible with the JPEG standard, which requires transmission of the quantization matrix.

Abstract: Image compression techniques based on fractals have been developed in the last few years and may promise better compression performance. Fractal image compression techniques are being developed due to the recognition that fractals can describe natural scenes better than shapes of traditional geometry. This paper describes principle and common techniques of fractal image compression. Mathematical foundations for fract image compression techniques are presented first. Then three main fractal image compression techniques are discussed. The first and most important technique is based on iterated function systems (IFS): images are compressed into compact IFS codes at encoding stage, and fractal images are generated to approximate the original image at the decoding stage. The second technique is segment-based coding: images are segmented according to the fractal dimension and these segments are coded efficiently using properties of the human visual system. The third technique is yardstick coding which is similar to DPCM and subsampling with subsequent interpolation. But it was inspired by the fractal geometry on measuring the length of a curve using a yardstick.

Abstract: The applications of discrete cosine transform (DCT)-based image- and video-coding methods in the asynchronous transfer mode (ATM) environment are considered. Coding and reconstruction mechanisms are jointly designed to achieve a good compromise among compression gain, system complexity, processing delay, error-concealment capability, and reconstruction quality. The Joint Photographic Experts Group (JPEG) and Motion Picture Experts Group (MPEG) algorithms for image and video compression are modified to incorporate block interleaving in the spatial domain and DCT coefficient segmentation in the frequency domain to conceal the errors due to packet loss. A new algorithm is developed that recovers the damaged regions by adaptive interpolation in the spatial, temporal, and frequency domains. The weights used for spatial and temporal interpolations are varied according to the motion content and loss patterns of the damaged regions. When combined with proper layered transmission, the proposed coding and reconstruction methods can handle very high packet-loss rates at only a slight cost in compression gain, system complexity, and processing delay. >

Abstract: An important issue that affects response time performance in current OODB and hypertext systems is the I/O involved in moving objects from slow memory to cache. A promising way to tackle this problem is to use prefetching, in which we predict the user's next page requests and get those pages into cache in the background. Current databases perform limited prefetching using techniques derived from older virtual memory systems. A novel idea of using data compression techniques for prefetching was recently advocated in [KrV, ViK], in which prefetchers based on the Lempel-Ziv data compressor (the UNIX compress command) were shown theoretically to be optimal in the limit. In this paper we analyze the practical aspects of using data compression techniques for prefetching. We adapt three well-known data compressors to get three simple, deterministic, and universal prefetchers. We simulate our prefetchers on sequences of page accesses derived from the OO1 and OO7 benchmarks and from CAD applications, and demonstrate significant reductions in fault-rate. We examine the important issues of cache replacement, size of the data structure used by the prefetcher, and problems arising from bursts of “fast” page requests (that leave virtually no time between adjacent requests for prefetching and book keeping). We conclude that prediction for prefetching based on data compression techniques holds great promise.

Abstract: A new algorithm for ECG signal compression is introduced. The compression system is based on the subautoregression (SAR) model, known also as the long-term prediction (LTP) model. The periodicity of the ECG signal is employed in order to further reduce redundancy, thus yielding high compression ratios. The suggested algorithm was evaluated using an in-house database. Very low bit rates on the order of 70 b/s are achieved with a relatively low reconstruction error (percent RMS difference-PRD) of less than 10%. The algorithm was compared, using the same database, with the conventional linear prediction (short-term prediction-STP) method, and was found superior at any bit rate. The suggested algorithm can be considered a generalization of the recently published average beat subtraction method. >

Abstract: The authors propose a lossless algorithm based on regularities, such as the presence of palindromes, in the DNA. The results obtained, although not satisfactory, are far beyond classical algorithms. >

Abstract: A family of algorithms that implement operations on compressed digital images is described. These algorithms allow many traditional image manipulation operations to be performed 50 to 100 times faster than their brute-force counterparts. It is shown how the algebraic operations of pixel-wise and scalar addition and multiplication, which are the basis for many image transformations, can be implemented on compressed images. These operations are used to implement two common video transformations: dissolving one video sequence into another and subtitling. The performance of these operations is compared with the brute-force approach. The limitations of the technique, extensions to other compression standards and the relationship of this research to other work in the area are discussed. >

Abstract: A video compression system comprises a pre-processing section (102), and encoder (106), and post-processing section (114). The pre-processing section (102) employs a median decimation filter (122) which combines median filtering and decimation process. The pre-processing section (102) also employs adaptive temporal filtering and content adaptive noise reduction filtering to provide images with proper smoothness and sharpness to match the encoder characteristics. The encoder (106) employs a two pass look-ahead allocation rate buffer control scheme where the numbers of bits allocated and subsequently generated for each block may differ. In the first pass, the means square error for each block is estimated to determine the number of bits assigned to each block in a frame. In the second pass, the degree of compression is controlled as a function of the total number of bits generated for all the preceding blocks and the sum of the bits allocated to such preceding blocks.

Abstract: The zero-tree method for image compression, proposed by J. Shapiro (1992), is studied. The method is presented in a more general perspective, so that its characteristics can be better understood. From this analysis, an improved method is proposed, and it is shown that the new method can increase the PSNR up to 1.3 dB over the original method. >

Abstract: Preface. 1. Hierarchical Model-Based Motion Estimation P. Anandan, J.R. Bergen, K.J. Hanna, R. Hingorani. 2. An Estimation Theoretic Perspective on Image Processing and the Calculation of Optical Flow T.M. Chin, M.R. Luettgen, W.C. Karl, A.S. Willsky. 3. Estimation of 2-D Motion Fields from Image Sequences with Application to Motion-Compensated Processing J. Konrad, E. Dubois. 4. Edge-Based 3-D Camera Motion Estimation with Application to Video Coding E. Zakhor, F. Lari. 5. Motion Compensation: Visual Aspects, Accuracy, and Fundamental Limits B. Girod. 6. Motion Field Estimators and their Application to Image Interpolation S. Tubaro, F. Rocca. 7. Subsampling of Digital Image Sequences using Motion Information R.A.F. Belfor, R.L. Lagendijk, J. Biemond. 8. Image Sequence Coding using Motion-Compensated Subband Decomposition A. Nicoulin, M. Mattavelli, W. Li, A. Basso, A. Popat, M. Kunt. 9. Vector Quantization for Video Data Compression R.M. Mersereau, M.J.T. Smith, C.S. Kim, F. Kossentini, K.K. Truong. 10. Model-Based Image Sequence Coding M. Buck, N. Diehl. 11. Human Facial Motion Analysis and Synthesis with Applications to Model-Based Coding K. Aizawa, C.s. Choi, H. Harashima, T.S. Huang. 12. Motion Compensated Spatiotemporal Kalman Filtering J.W. Woods, J. Kim. 13. Multiframe Wiener Restoration of Image Sequences M.K. Ozkan, M.I. Sezan, A.T. Erdem, A.M. Tekalp. 14. 3-D Median Structures for Image Sequence Filtering and Coding T. Viero, Y. Neuvo. 15. Video Compression for Digital ATV Systems J.G. Apostolopoulos, J.S. Lim. Index.

Abstract: A novel variable-size block multi-resolution motion estimation (MRME) scheme is presented. The motion estimation scheme can be used to estimate motion vectors in subband coding, wavelet coding and other pyramid coding systems for video compression. In the MRME scheme, the motion vectors in the highest layer of the pyramid are first estimated, then these motion vectors are used as the initial estimate for the next layer and gradually refined. A variable block size is used to adapt to its level in the pyramid. This scheme not only considerably reduces the searching and matching time but also provides a meaningful characterization of the intrinsic motion structure. In addition, the variable-MRME approach avoids the drawback of the constant-size MRME in describing small object motion activities. The proposed variable-block size MRME scheme can be used in estimating motion vectors for different video source formats and resolutions including video telephone, NTSC/PAL/SECAM, and HDTV applications.

Abstract: A data compression system greatly compresses the stored data used by a speech recognition system employing hidden Markov models (HMM). The speech recognition system vector quantizes the acoustic space spoken by humans by dividing it into a predetermined number of acoustic features that are stored as codewords in a vector quantization (output probability) table or codebook. For each spoken word, the speech recognition system calculates an output probability value for each codeword, the output probability value representing an estimated probability that the word will be spoken using the acoustic feature associated with the codeword. The probability values are stored in an output probability table indexed by each codeword and by each word in a vocabulary. The output probability table is arranged to allow compression of the probability of values associated with each codeword based on other probability values associated with the same codeword, thereby compressing the stored output probability. By compressing the probability values associated with each codeword separate from the probability values associated with other codewords, the speech recognition system can recognize spoken words without having to decompress the entire output probability table. In a preferred embodiment, additional compression is achieved by quantizing the probability values into 16 buckets with an equal number of probability values in each bucket. By quantizing the probability values into buckets, additional redundancy is added to the output probability table, which allows the output probability table to be additionally compressed.

Abstract: Most image or video applications involving transmission or storage require some form of data compression to reduce the otherwise inordinate demand on bandwidth and storage. Compatibility among different applications and manufacturers is very desirable, and often essential. This paper describes several standard compression algorithms developed in recent years.

Abstract: A method for compressing video movie data to a specified target size using intraframe and interframe compression schemes. In intraframe compression, a frame of the movie is compressed by comparing adjacent pixels within the same frame. In contrast, interframe compression compresses by comparing similarly situated pixels of adjacent frames. The method begins by compressing the first frame of the video movie using intraframe compression. The first stage of the intraframe compression process does not degrade the quality of the original data, e.g., the method uses run length encoding based on the pixels' color values to compress the video data. However, in circumstances where lossless compression is not sufficient, the method utilizes a threshold value, or tolerance, to achieve further compression. In these cases, if the color variance between pixels is less than or equal to the tolerance, the method will encode the two pixels using a single color value--otherwise, the method will encode the two pixels using different color values. The method increases or decreases the tolerance to achieve compression within the target range. In cases where compression within the target range results in an image of unacceptable quality, the method will split the raw data in half and compress each portion of data separately. Frames after the first frame are generally compressed using a combination of intraframe and interframe compression. Additionally, the method periodically encodes frames using intraframe compression only in order to enhance random frame access.

Abstract: An improved electronic solid-state record/playback device (SSRPD) and electronic system may be used to record and playback information such as audio, video, control, and other data. The SSRPD uses no tape or moving parts in the actual record/playback process but includes an audio and/or video and/or other data record/playback module (RPM), which performs all of the record signal conversion, recording and data compression algorithms, digital signal processing, and playback signal conversion. The SSRPD has program input processing and control output processing modules so that other devices may be controlled in different ways including interactive control. A time and control processor module facilitates internal synchronization of the SSRPD audio, video, and control information, as well as synchronization with other devices. The SSRPD information described is recorded into an internal resident memory(s). The novel interface allows information to be exchanged without degradation via a digital Portable Storage Device (PSD) which may be a Random Access Memory card (RAM card), with other SSRPDs as well as to a special Computer Interface Device (CID). The CID is an intelligent device that connects to a standard computing device such as a PC and facilitates functions such as reading, writing, editing, and archiving PSD data, as well as performing diagnostic routines.

Abstract: A method for temporal prediction of image sequences is proposed. The motion vectors of conventional block-based motion compensation schemes are used to convey a mapping of a selected set of image points, instead of blocks, between the previous and the current image. The prediction is made by geometrically transforming, or warping, the previous image using the point pairs defined by the mapping as fixed points in the transformation. This method produces a prediction image without block artifacts and can compensate for many motion types where conventional block matching fails, such as scaling and rotation. It can also be incorporated in existing hybrid video compression systems with little additional complexity and few or no changes in the bit stream syntax. It is shown that a significant subjective improvement in the prediction as well as a consistent reduction in the objectively measured prediction error is obtained. >

Abstract: A review is presented of vector quantization, the mapping of pixel intensity vectors into binary vectors indexing a limited number of possible reproductions, which is a popular image compression algorithm. Compression has traditionally been done with little regard for image processing operations that may precede or follow the compression step. Recent work has used vector quantization both to simplify image processing tasks, such as enhancement classification, halftoning, and edge detection, and to reduce the computational complexity by performing the tasks simultaneously with the compression. The fundamental ideas of vector quantization are explained, and vector quantization algorithms that perform image processing are surveyed. >

Abstract: A video coding scheme based on wavelet representation performs motion compensation in the wavelet domain rather than spatial domain. This inter-frame wavelet transform coding scheme preferably uses a variable-size multi-resolution motion compensation (MRMC) prediction scheme. The MRMC scheme produces displaced residual wavelets (DRWs). An optimal bit allocation algorithm produces a bit map for each DRW, and each DRW is then adaptively quantized with its respective bit map. Each quantized DRW is then coded into a bit stream.

Abstract: Packet switched communications services with real-time delay constraints, such as voice and video, combine the established fields of digital signal processing and data communications networking. Each field is outlined, and new open problems due to the combination are identified. A simulation study investigates layered coding using a packet voice Markov chain source model. Information is partitioned into two or more priority layers to protect important components from loss. A parameter, $\alpha$, identifies the proportion of traffic placed in each priority. Packet loss rates for each priority and $\alpha$ are used to compute the signal to noise ratio, which is a more appropriate performance measure for voice and video services than loss rates alone. Dynamic feedback control of the priority partition based on network load conditions is shown to be effective, even with substantial feedback delay. Feedback, in conjunction with priority, provides graceful service degradation with increasing load and loss rate. A queueing analysis of this system is also investigated using a two-dimensional Markov chain source model that represents both load and $\alpha$, which vary dynamically. The simulation and analytic models are compared. Two methods for reducing the numerical complexity are given. A two-hour long empirical sample of variable rate video is derived by applying a simple intraframe video compression code to an action movie. Statistical characteristics are measured, including an accurate model for the heavy-tailed marginal distribution of video frame bandwidth. A statistical property called long-range dependence is described, measured, and shown to be significant for this data. A new traffic source model results from combining the marginal distribution with long-range dependence. Extensive trace driven simulations characterize network queueing behavior and allocation of bandwidth/buffer resources. Statistical multiplexing gain of variable rate video is evaluated as well as the advantage due to multiplexing video with data services. We discuss the implications of this traffic analysis for for the design of congestion control mechanisms for integrated packet networks. We close with some comments on the ramifications of advancing electronic hardware speed and complexity for multi-media communications.

Abstract: A class of lossless data compression algorithms use a memory-based dictionary of finite size to facilitate the compression and decompression of data. To reduce the loss in data compression caused by dictionary resets, a standby dictionary is used to store a subset of encoded data entries previously stored in a current dictionary. In a second aspect of the invention, data is compressed/decompressed according to the address location of data entries contained within a dictionary built in a content addressable memory (CAM). In a third aspect of the invention, the minimum memory/high compression capacity of the standby dictionary scheme is combined with the fast single-cycle per character encoding/decoding capacity of the CAM circuit. The circuit uses multiple dictionaries within the storage locations of a CAM to reduce the amount of memory required to provide a high data compression ratio.

Abstract: A motion-adaptive variable-bit-rate (VBR) video codec is considered, and a motion-classified model is developed to represent the characteristics of various classes of motion activities, including scene changes. The codec switches between interframe, motion-compensated, and intraframe coding corresponding to low, medium, and high amounts of motion and scene changes, respectively. The model captures the motion of various video scenes by providing the statistics of VBR-coded video traffic through a first-order autoregressive process with time-varying parameters. The parameters of this model are obtained from a VBR-coded sample video sequence with the objective of matching the bit-rate distribution and the autocorrelation among the bit rates. The validity and accuracy of the model are evaluated, and the characteristics of aggregated traffic sources obtained with the model are discussed. >

Abstract: A variable-bit-rate (VBR) MPEG video compression encoder is introduced, and the performance of a statistically multiplexed asynchronous transfer mode (ATM) network supporting a number of such VBR video sources is evaluated. Bit-rate characteristics obtained from a detailed simulation are provided for a VBR MPEG encoder for CCIR601 video (operating in the 5-10 Mb/s regime) appropriate for medium-quality multimedia or broadcasting applications. The results presented include bit-rate traces and signal-to-noise-ratio data for typical test sequences, along with summary statistics such as the marginal distribution of frame rate. Data from a study of statistical multiplexing on an ATM network are also given. Simulation results for an ATM statistical multiplexer with N>>1 VBR MPEG sources are presented in terms of key performance measures such as cell loss rate and delay versus throughput. The results confirm that ATM channel efficiencies of approximately 80-90% can be obtained at reasonable cell loss rate and delay levels. >

Abstract: A novel decoding algorithm for the MC-DCT (motion-compensation discrete-cosine-transform)-based video, which performs inverse MC before inverse DCT, is designed. This algorithm can be applied in compositing compressed video within the network, which may take multiple compressed video sources and combine them into a single compressed output stream. The proposed algorithm convers all MC-DCT compressed video into the DCT domain and performs compositing in the DCT domain. This DCT-domain approach can reduce the required computations with a speedup factor depending on the compression ratio and the nonzero motion vector percentage. However, dropping some least-significant DCT coefficients may be necessary for the worst case of high-motion video in real-time implementations. Some issues of networked video compositing are also discussed. Another direct application of the proposed decoding algorithm is converting MC-DCT compressed video to the DCT compressed format directly in the DCT domain. >

Abstract: Wavelet image decompositions generate a tree-structured set of coefficients, providing an hierarchical data-structure for representing images. While early wavelet-based algorithms for image compression concentrated on optimal quantization of wavelet coefficients, several recent researchers have proposed approaches which couple coefficient quantization (either scalar or vector-based) with various strategies for quantizing the tree itself. This paper proposes an image compression algorithm based on optimal bit rate allocation between scalar and tree-structured quantizers. A predictive approach to representing the pruned tree structure is presented, and the entropy of this representation is included in the optimal allocation problem. The algorithm couples Lagrangian optimization of scalar quantizers with a marginal analysis approach for optimizing the tree structure, and achieves excellent coding efficiency in the rate-distortion sense. >

Abstract: A data compression/decompression processor (a single-chip VLSI data compression/decompression engine) for use in applications including but not limited to data storage and communications The processor is highly versatile such that it can be used on a host bus or housed in host adapters, so that all devices such as magnetic disks, tape drives, optical drives and the like connected to it can have substantial expanded capacity and/or higher data transfer rate The processor employs an advanced adaptive data compression algorithm with string-matching and link-list techniques so that it is completely adaptive, and a dictionary is constructed on the fly No prior knowledge of the statistics of the characters in the data is needed During decompression, the dictionary is reconstructed at the same time as the decoding occurs The compression converges very quickly and the compression ratio approaches the theoretical limit The processor is also insensitive to error propagation

Abstract: Uses an algorithm based on the adapted-window Malvar transform to decompose digitized speech signals into a local time-frequency representation. The authors present some applications and experimental results for a signal compression and automatic voiced-unvoiced segmentation. This decomposition provides a method of parameter simplification which appears to be useful for detecting fundamental frequencies, and characterizing formants. >

Abstract: An approach to the lossy compression of color images with limited palette that does not require color quantization of the decoded image is presented. The algorithm is particularly suited for coding images using an image-dependent palette. The technique restricts the pixels of the decoded image to take values only in the original palette. Thus, the decoded image can be readily displayed without having to be quantized. For comparable quality and bit rates, the technique significantly reduces the decoder computational complexity. >