Date 2830 March 1995
Filter Results

Proceedings DCC '95 Data Compression Conference [table of contents]
Publication Year: 1995 
Quantization distortion in block transformcompressed data
Publication Year: 1995
Cited by: Papers (1)  Patents (8)Summary form only given, as follows. The JPEG image compression standard is an example of a block transformbased compression scheme; the image is systematically subdivided into blocks that are individually transformed, quantized, and encoded. The compression is achieved by quantizing the transformed data, reducing the data entropy and thus facilitating efficient encoding. Block transform compression schemes exhibit sharp discontinuities at data block boundaries: this phenomenon is a visible manifestation of the compression quantization distortion. For example, in compression algorithms such as JPEG these blocking effects manifest themselves visually as discontinuities between adjacent 8×8 pixel image blocks. In general the distortion characteristics of block transformbased compression techniques are understandable in terms of the properties of the transform basis functions and the transform coefficient quantization error. In particular, the blocking effects exhibited by JPEG are explained by two simple observations demonstrated in this work: a disproportionate fraction of the total quantization error accumulates on block edge pixels; and the quantization errors among pixels within a compression block are highly correlated, while the quantization errors between pixels in separate blocks are uncorrelated. A generic model of block transform compression quantization noise is introduced, applied to synthesized and real one and two dimensional data using the DCT as the transform basis, and results of the model are shown to predict distortion patterns observed in data compressed with JPEG. View full abstract»

Efficient handling of large sets of tuples with sharing trees
Publication Year: 1995
Cited by: Papers (1)Summary form only given; substantially as follows. Computing with sets of tuples (nary relations) is often required in programming, while being a major cause of performance degradation as the size of sets increases. The authors present a new data structure dedicated to the manipulation of large sets of tuples, dubbed a sharing tree. The main idea to reduce memory consumption is to share some subtuples of the set represented by a sharing tree. Various conditions are given. The authors have developed algorithms for common set operations: membership, insertion, equality, union, intersection, ... that have theoretical complexities proportional to the sizes of the sharing trees given as arguments, which are usually much smaller than the sizes of the represented sets. View full abstract»

PPT based fast architecture & algorithm for discrete wavelet transforms
Publication Year: 1995First Page of the ArticleView full abstract» 
Algorithm evaluation for the synchronous data compression standards
Publication Year: 1995First Page of the ArticleView full abstract» 
An investigation of effective compression ratios for the proposed synchronous data compression proto
Publication Year: 1995First Page of the ArticleView full abstract» 
The development of a standard for compression of synchronous data in DSU/CSU's
Publication Year: 1995Summary form only given, as follows. Over the past year, a standard for compression of synchronous data in DSU/CSUs (56 kb/s) has been developed. The development began in an informal industry consortium known as the Synchronous Data Compression Consortium (SDCC) and the work later migrated to a committee of the Telecommunications Industry Association (TIA). This work chronicles the development of the standard, which is based on the Internet Standard Pointtopoint Protocol, examining both the issues involved when applying data compression to communication links and the impact of choices made along the development path. View full abstract»

Parallel image compression using vector quantization
Publication Year: 1995Summary form only given. The authors used a parallel approach to address the complexity issues of vector quantization: They implemented two full search memoryless parallel vector quantizers using 2 x 2 and 4 x 4 fixed block sizes on a shared memory MIMD machine, the BBN GP 1000. The squared error distortion measure and the LBG codebook design algorithm were used. The searching of the codebook is done in parallel for both the image coding and codebook design phases. The input vectors are in shared memory distributed among all the processor node memory modules. A private copy of the codebook is given to each processor node. A parallel task is generated for each input vector to be encoded. Each task is assigned one input vector to encode. The task searches the entire codebook to determine the minimum distortion codevector. The index of this vector is the output of the task. Load balancing of the tasks on the available processor nodes is done automatically by the operating system. This algorithm design requires minimal synchronization between the tasks to accumulate the total distortion. While good parallel performance was achieved the vector quantizers were generally lacking in fidelity. It is expected that these methods can be extended to achiwe high fidelity nhile maintaining good parallel performance. View full abstract»

Making the Best of JPEG at Very High Compression Ratios: Rectangular Pixel Averaging for Mars PathFinder
Publication Year: 1995As more and more NASA missions are turning to image compression to maximize their data return at constrained bit rates, and very often adopting JPEG as the centerpiece of their image compression system, they are noticing one limitation of JPEG: its poor performance at very high compression ratios (typically 32 and above). This is the case for engineering uses of image data on Mars Pathfinder such as assessment of lander condition and deployed airbags, and rover navigation. Unlike science scenes, images for engineering at low resolution are often sufficient. Unfortunately, at very high compression ratios, JPEG produces unacceptable artifacts, due to the limitation of the size of the Discrete Cosine Transform to 8, for which no clever quantization or entropy coding can compensate. Still, JPEG can successfully be used as part of a compression scheme if the encoding is preceeded by lowpass filtering and downsampling, while the decoding is followed by interpolation and upsampling, to restore the image to its original size. The choice for the horizontal and vertical downsampling/upsampling factors is made based on the known distance to the objective and its size, as well as on the fact that resolution in azimuth degrades more gradually than in elevation, leading to a larger downsampling factor in azimuth. Assuming unweighted pixel averaging is used as the lowpass filter before decimation, optimal interpolation filters which minimize the mean squared reconstruction error (MSRE) in the absence of JPEG are derived. In the presence of large JPEGinduced quantization noise however, bilinear interpolation filters are shown to outperform the optimal interpolation filters derived above. Engineering assessment of images compressed with this scheme at ratios up to 126to1 using bilinear interpolation confirm its performance and its success in extending the operational compression ratio range of JPEG. View full abstract»

The Implementation of Data Compression in the Cassini RPWS Dedicated Compression Processor
Publication Year: 1995
Cited by: Papers (1)First Page of the ArticleView full abstract» 
Enhancement of IMP lossy image data compression using LCT
Publication Year: 1995
Cited by: Papers (1)Summary form only given, as follows. This work addresses the lossy image data compression designed for the NASA Mars pathfinder (IMP) project. Both, mission profile and the availability of a RISC central board computer support a completely software oriented implementation of IMP lossy image data compression. One of the mission objectives is to demontfate the capability to place science payloads on the surface of Mars using a simple, reliable, low cost system within a demanding schedule. According to the cost and schedule objective a task oriented modification of the widely used Joint Photographic Expeqt Group (JPEG) standard for still image data compression was implemented as a starting version of compression software. This version has been delivered within a month after project kickoff. Subsequently extensions and improvements of performance were investigated and implemented as : (a) selection between pixel representation by 8 or 12 bit, 'item[(b)] implementation of arithmetic coding (improvement of performance), (c) implementation of "local cosine transform" (improvement of performance). View full abstract»

Lossless compression using conditional entropyconstrained subband quantization
Publication Year: 1995
Cited by: Papers (1)Summary form only given, as follows. The browse and residual compression strategy has been shown to be effective for data archival and telebrowsing of scientific databases. This paper introduces a hybrid lossless image compression technique that couples lossy subband quantization with lossless coding of the residual. Analysis is performed on the tradeoff between the rates expended on the browse and residual images. We investigate the effects of different distortion measures for compression of the browse image on the compression of the residue. The algorithm is shown to provide competitive lossless compression as well as the flexibility for progressive transmission for a moderate computational complexity. View full abstract»

A massively parallel algorithm for vector quantization
Publication Year: 1995
Cited by: Papers (2)Summary form only given, as follows. This work is concerned with the parallel implementation of a vector quantizer system on Maspar MP2, a singleinstruction, multipledata (SIMD) massively parallel computer. A vector quantizer (VQ) consists of two mappings: an encoder and a decoder. The encoder assigns to each input vector the index of the codevector that is closest to it. The decoder uses this index to reconstruct the signal. In our work, we used the Euclidean distortion measure to find the codevector closest to each input vector. The work described in this paper used a Maspar MP2216 located at the Goddard Space Flight Center, Greenbelt, Maryland. This system has 16,384 processor elements (PEs) arranged in a rectangular array of 128 x 128 nodes. The parallel VQ algorithm is based on pipelining. The codevectors are distributed equally among the PEs in the first row of the PE array. These codevectors are then duplicated on the remaining processor rows. Traversing along any row of the PE array amounts to traversing through the entire codebook. After populating the PEs with the codevectors, the input vectors are presented to the first column of PEs. Each PE receives one vector at a time. The first set of data vectors are now compared with the group of codevectors in the first column. A data packet containing the the input vector, the minimum value of the distortion between the input vector and the code vectors it has encountered so far, and the index corresponding to the codevector that accounted for the current minimum value of the distortion is associated with each input vector. After updating the entries of the data packet, it is shifted one column to the right in the PE array. The next set of input vectors takes its place in the first column. The above process is repeated till all the input vectors are exhausted. The indices for the first set of data vectors are obtained after an appropriate number of shifts. The remaining indices are obtained in subsequent shi ts. Results of extensive performance evaluations are presented in the fulllength paper. These results suggest that our algorithm makes very efficient use of the parallel capabilities of the Maspar system. The existence of efficient algorithms such as the one presented in this paper should increase the usefulness and applicability of vector quantizers in Earth and Space science applications. View full abstract»

Author index
Publication Year: 1995 
Coding gain of intra/interframe subband systems
Publication Year: 1995
Cited by: Papers (1)Summary form only given. Typical image sequence coders use motion compensation techniques in connection to coding of the motion compensated difference images (interframe coding). Moreover, the difference loop is initialized from time to time by intraframe coding of images. It is therefore important to have a procedure that allows to evaluate the performance of a particular coding scheme: coding gain and ratedistortion figures are used in this work to this purpose. We present an explicit procedure to compute the coding gain for twodimensional separable subband systems, both in the case of a uniform and a pyramid subband decomposition, and for the case of interframe coding. The technique operates in the signal domain and requires the knowledge of the autocorrelation function of the input process. In the case of a separable subband system and image spectrum, the coding gain can be computed by combining the results relative to appropriately defined onedimensional filtering schemes, thus making the technique very attractive in terms of computational complexity. We consider both the case of a uniform subband decomposition and of a pyramid decomposition. The developed procedure is applied to compute the subband coding gain for motion compensated signals in the case of images modeled as separable Markov processes: different filter banks are compared to each other and to transform coding. In order to have indications on the effectiveness of motion compensation, we also compute the coding gain for intraframe images. We show that the results for the image models are in very good agreement with those obtained with realworld data View full abstract»

Recursively indexed vector quantization of nonstationary sources
Publication Year: 1995
Cited by: Papers (2)Summary form only given. We present a recursively indexed vector quantizer with the following properties: (1) it is simple to implement with low computational overhead; (2) it is an adaptive algorithm and therefore is well suited for applications where the source is nonstationary; (3) the output rate can easily be changed making it suitable for applications requiring rate control, such as transmission over packet switched networks; and (4) the input vectors can be quantized to within a user specified distortion on a per vector basis rather than on average. We have called the algorithm forward adaptive even though this algorithm also use the past outputs for adaptation. We have tested this algorithm on a number of synthetic hidden Markov sources, and on a video sequence. The results of both tests compare favorably with existing results in the literature View full abstract»

Video coding using 3 dimensional DCT and dynamic code selection
Publication Year: 1995
Cited by: Papers (1)Summary only given. We address the quality issue, and present a method for improved coding of the 3D DCT coefficients. Performance gain is achieved through the use of dynamically selected multiple coding algorithms. The resulting performance is excellent giving a compression ratio of greater than to 100:1 for image reproduction. The process consists of stacking 8 frames and breaking the data into 8×8×8 pixel cubes. The three dimensional DCT is applied to each cube. Each cube is then scanned in each dimension to determine if significant energy exists beyond the first two coefficients. Significance is determined with separate thresholds for each dimension. A single bit of side information is transmitted for each dimension of each cube to indicate whether more than two coefficients will be transmitted. The remaining coefficients of all cubes are reordered into a linear array such that the elements with the highest expected energies appear first and lower expected energies appear last. This tends to group coefficients with similar statistical properties for the most efficient coding. Eight different encoding methods are used to convert the coefficients into bits for transmission. The Viterbi algorithm is used to select the best coding method. The cost function is the number of bits that need to be sent. Each of the eight coding methods is optimized for a different range of values View full abstract»

Constraining the size of the instantaneous alphabet in trellis quantizers
Publication Year: 1995 , Page(s): 23  32A method is developed for decreasing the computational complexity of a trellis quantizer (TQ) encoder. We begin by developing a ratedistortion theory under a constraint on the average instantaneous number of quanta considered. This constraint has practical importance: in a TQ, the average instantaneous number of quanta is exactly the average number of multiplies required at the encoder. The theory shows that if the conditional probability of each quanta is restricted to a finite region of support, the instantaneous number of quanta considered can be made quite small at little or no cost in SQNR performance. Simulations of TQs confirm this prediction. This reduction in complexity makes practical the use of modelbased TQs (MTQs), which had previously been considered computationally unreasonable. For speech, performance gains of several dB SQNR over adaptive predictive schemes at a similar computational complexity are obtained using only a firstorder MTQ View full abstract»

Treestructured vector quantization with significance map for wavelet image coding
Publication Year: 1995 , Page(s): 33  41
Cited by: Papers (8)  Patents (2)Variablerate treestructured VQ is applied to the coefficients obtained from an orthogonal wavelet decomposition. After encoding a vector, we examine the spatially corresponding vectors in the higher subbands to see whether or not they are “significant”, that is, above some threshold. One bit of side information is sent to the decoder to inform it of the result. When the higher bands are encoded, those vectors which were earlier marked as insignificant are not coded. An improved version of the algorithm makes the decision not to code vectors from the higher bands based on a distortion/rate tradeoff rather than a strict thresholding criterion. Results of this method on the test image “Lena” yielded a PSNR of 30.15 dB at 0.174 bits per pixel View full abstract»

Matching pursuit video coding at very low bit rates
Publication Year: 1995 , Page(s): 411  420
Cited by: Papers (8)  Patents (10)Matching pursuits refers to a greedy algorithm which matches structures in a signal to a large dictionary of functions. In this paper, we present a matchingpursuit based video coding system which codes motion residual images using a large dictionary of Gabor functions. One feature of our system is that bits are assigned progressively to the highestenergy areas in the motion residual image. The large dictionary size is another advantage, since it allows structures in the motion residual to be represented using few significant coefficients. Experimental results compare the performance of the matchingpursuit system to a hybridDCT system at various bit rates between 6 and 128 kbit/s. Additional experiments show how the matching pursuit system performs if the Gabor dictionary is replaced by an 8×8 DCT dictionary View full abstract»

Vector quantization for lossless textual data compression
Publication Year: 1995
Cited by: Patents (1)Summary form only given. Vector quantisation (VQ) may be adapted for lossless data compression if the data exhibit vector structures, such as in textural relational databases. Lossless VQ is discussed and it is demonstrated that a relation of tuples may be encoded and allocated to physical disk blocks such that standard database operations such as access, insertion, deletion, and update may be fully supported View full abstract»

A high performance block compression algorithm for small systemssoftware and hardware implementations
Publication Year: 1995Summary form only given. A new algorithmic approach to block data compression, using a highly contextual codification of the dictionary, that gives substantial compressionrate advantages over existing technologies, is described. The algorithm takes into account the limitations and characteristics of small systems, such as a low consumption of memory, high speed and short latency, as required by communication applications. It uses a novel construction of the prefixfree dictionary, a simple but powerful heuristic for filtering out the noncompressed symbols and a predictive dynamic prefix coding for the output entities. It also employs universal codification of the integers, allowing a very fast and direct implementation in silicon. A dynamic compression software package is detailed. Also, several techniques developed to maximize the usable diskspace and the software speed, among others, are discussed View full abstract»

Latticebased designs of direct sum codebooks for vector quantization
Publication Year: 1995
Cited by: Papers (1)Summary form only given. A direct sum codebook (DSC) has the potential to reduce both memory and computational costs of vector quantization. A DSC consists of several sets or stages of vectors. An equivalent code vector is made from the direct sum of one vector from each stage. Such a structure, with p stages containing m vectors each, has m^{p} equivalent code vectors, while requiring the storage of only mp vectors. DSC quantizers are not only memory efficient, they also have a naturally simple encoding algorithm, called a residual encoding. A residual encoding uses the nearest neighbor at each stage, requiring comparison with mp vectors rather than all m^{p} possible combinations. Unfortunately, this encoding algorithm is suboptimal because of a problem called entanglement. Entanglement occurs when a different vector from that obtained by a residual encoding is actually a better fit for the input vector. An optimal encoding can be obtained by an exhaustive search, but this sacrifices the savings in computation. Latticebased DSC quantizers are designed to be optimal under a residual encoding by avoiding entanglement Successive stages of the codebook produce finer and finer partitions of the space, resulting in equivalent code vectors which are points in a truncated lattice. After the initial design, the codebook can be optimized for a given source, increasing performance beyond that of a simple lattice vector quantizer. Experimental results show that DSC quantizers based on cubical lattices perform as well as exhaustive search quantizers on a scalar source View full abstract»

Parallel algorithms for the static dictionary compression
Publication Year: 1995 , Page(s): 162  171
Cited by: Papers (6)Studies parallel algorithms for two static dictionary compression strategies. One is the optimal dictionary compression with dictionaries that have the prefix property, for which our algorithm requires O(L+log n) time and O(n) processors, where L is the maximum allowable length of the dictionary entries, while previous results run in O(L+log n) time using O(n^{2}) processors, or in O(L+log^{2}n) time using O(n) processors. The other algorithm is the longestfragmentfirst (LFF) dictionary compression, for which our algorithm requires O(L+log n) time and O(nL) processors, while the previous result has O(L log n) time performance on O(n/log n) processors. We also show that the sequential LFF dictionary compression can be computed online with a lookahead of length O(L^{2}) View full abstract»

Improving LZFG data compression algorithm
Publication Year: 1995Summary form only given. This paper presents two approaches to improve the LZFG data compression algorithm. One approach is to introduce a selfadaptive word based scheme to achieve significant improvement for English text compression. The other is to apply a simple movetofront scheme to further reduce the redundancy within the statistics of copy nodes. The experiments show that an overall improvement is achieved from both approaches. The selfadaptive wordbased scheme takes all the consecutive English characters as one word. Any other character in the ASCII codes will be taken as one single word. As an example, the input message `(2+x) is represented by y' can be classified into 9 words. To run the wordbased scheme in PATRICIA tree, the data structure is modified View full abstract»