We introduce a new method, called Tree K-SVD, to learn a tree-structured dictionary for sparse representations, as well as a new adaptive sparse coding method, in a context of image compression. Each dictionary at a level in the tree is learned from residuals from the previous level with the K-SVD method. The tree-structured dictionary allows efficient search of the atoms along the tree as well as efficient coding of their indices. Besides, it is scalable in the sense that it can be used, once learned, for several sparsity constraints. We show experimentally on face images that, for a high sparsity, Tree K-SVD offers better rate-distortion performances than state-of-the-art “flat” dictionaries learned by K-SVD or Sparse K-SVD, or than the predetermined overcomplete DCT dictionary. We also show that our adaptive sparse coding method, used on a tree-structured dictionary to adapt the sparsity per level, improves the quality of reconstruction.

We introduce a new method to learn an adaptive dictionary structure suitable for efficient coding of sparse representations. The method is validated in a context of satellite image compression. The dictionary structure adapts itself during the learning to the training data and can be seen as a tree structure whose branches are progressively pruned depending on their usage rate and merged into a single branch. This adaptive structure allows a fast search for the atoms and an efficient coding of their indices. It is also scalable in sparsity, meaning that once learned, the structure can be used for several sparsity values. We show experimentally that this adaptive structure offers better rate-distortion performances than the “flat” K-SVD dictionary, a dictionary structured in one branch, and the tree-structured K-SVD dictionary (called Tree K-SVD).

We propose a new hierarchical approach to resolution scalable lossless and near-lossless (NLS) compression. It combines the adaptability of DPCM schemes with new hierarchical oriented predictors to provide resolution scalability with better compression performances than the usual hierarchical interpolation predictor or the wavelet transform. Because the proposed hierarchical oriented prediction (HOP) is not really efficient on smooth images, we also introduce new predictors, which are dynamically optimized using a least-square criterion. Lossless compression results, which are obtained on a large-scale medical image database, are more than 4% better on CTs and 9% better on MRIs than resolution scalable JPEG-2000 (J2K) and close to nonscalable CALIC. The HOP algorithm is also well suited for NLS compression, providing an interesting rate-distortion tradeoff compared with JPEG-LS and equivalent or a better PSNR than J2K for a high bit rate on noisy (native) medical images.

A new adaptive approach for lossless and near-lossless scalable compression of medical images is presented. It combines the adaptivity of DPCM schemes with hierarchical oriented prediction (HOP) in order to provide resolution scalability with better compression performances. We obtain lossless results which are about 4% better than resolution scalable JPEG2000 and close to non scalable CALIC on a large scale database. The HOP algorithm is also well suited for near-lossless compression, providing interesting rate-distortion trade-off compared to JPEG-LS and equivalent or better PSNR than JPEG2000 for high bit-rate on noisy (native) medical images.

This paper proposes to introduce Lanczos interpolation filters as wavelet atoms in an optimized decomposition for embedded lossy to lossless compression of biomedical images. The decomposition and the Lanczos parameter are jointly optimized in a generic packet structure in order to take into account the various contents of biomedical imaging modalities. Lossless experimental results are given on a large scale database. They show that in comparison with a well known basis using 5/3 biorthogonal wavelets and a dyadic decomposition, the proposed approach allows to improve the compression by more than 10% on less noisy images and up to 30% on 3D-MRI while providing similar results on noisy datasets.

The method we present is developed as part of an on-board satellite image compression with target ratios greater than six. In order to limit the ringing effects on these critical areas we have developed an adapted encoder based on a wavelet transform. Our method consists of first a classification of 8/spl times/8 blocks of coefficients and second of a coding which integrates a local quality constraint within a bit rate control. Actually, we perform coding and regulation in the same time by using an EZW encoder and also Shapiro's dependency protocol to define our blocks of coefficients, noted vectors. An energy-based threshold followed by an integration of the local context allows us to discriminate two classes of vectors. Once classification is set-up, we code the wavelet coefficients by using two EZW encoders, one for each class. Our encoder is quite different from Shapiro's because of the integration of a quality constraint. This constraint is established in order to privilege the ROI and is satisfied by an optimal choice of the proportionality between the two target rates.

This paper presents a method for video sequence coding, using motion estimation over a triangular object-based mesh. We introduce a novel "crack line" notion to deal with motion discontinuities and to improve representation scalability. The optimization of the nodal motion vectors is performed by means of a multi-resolution differential method, further improved by "backward in forward" tracking. To illustrate the proposed methods, results concerning the motion compensation of a mesh are reported.

In this paper, we present an efficient region-based video coding scheme. The region partioning is optimized according to a rate-distortion criterion in order to achieve constant bit-rate or constant quality. This optimization is led through a labeling for each image of an initial segmentation map using regions merging, regions' label relaxation and quantization variation. Results obtained show the performance of this coding scheme compared to MPEG and H263 codecs.

In the present paper we propose a new algorithm for planar facets segmentation of sequences of uncalibrated images in order to recover 3D models of complex scenes. This is performed using a two-step algorithm. First a robust and regularized dense disparity field is computed under the epipolar geometry constraint. The resulting field is segmented according to homographic models using an iterative Delaunay triangulation.

In this paper, we present an efficient region-based video coding scheme. The region partitioning is obtained using a minimum description length (MDL) formalism in order to optimize the compression. We compare this approach with results obtained by MPEG and H.261 coders and show the gain obtained by our approach.

In this paper we present a boundary-based representation of a motion-based image sequence segmentation. A tracking algorithm is associated to this structure which is able to handle several objects simultaneously, in the presence of partial occlusions. This segmentation representation uses the boundaries between all couples of adjacent regions approximated by a polygonal line. We extend it with a label on each boundary indicating to which region it is belonging from a motion point of view. This information is useful for an efficient predictive tracking of this representation over time. Moreover, we show that the structural information enclosed in our representation is very useful when regions are occluding. This representation can be used for multiple purposes in image sequence analysis (for example region fusion, layering) as well as coding and multimedia applications (for instance motion-compensated compression, temporal interpolation, object manipulation). Among these applications, we consider a temporal image sequence interpolation which can be done at the receiver end of a coder/decoder chain.

The number of regions and length of contour are two basic constraints in segmentation-based motion-compensated video coding. This paper presents a coding scheme which focuses on region number reduction, contour coding, and displaced frame difference (DFD) compression. One of the important features of the proposed scheme is a spatio-temporal simplification algorithm based on morphological filters, with which an image can be segmented into a small number of regions. Another important feature of the scheme is a segmentation map sampling technique which reduces contour length by about 50% with a very small reconstruction error. Experimental results show that, using the proposed scheme, a high compression ratio can be achieved with a small coding error for video sequences such as Miss America and Foreman.

This paper deals with new advances made in the field of discrete spatio-temporal filters applied to digital image sequences. Within time-varying images, the temporal correlation of the information is folded within the spatio-temporal domain by motions originating from both camera and object displacements. The spatio-temporal video information can be therefore reformulated in terms of motion trajectories and intensity variations along these trajectories. To yield that signal description, the spatio-temporal scenes will be segmented according to motion. Motion-compensated temporal filters have been used as convolutional filters applied along the assumed motion trajectories. Spectral interpretations show the efficiency of motion-compensated filtering for video signals. As a matter of fact, the whole signal analysis performed in this paper also includes a spatial filtering to achieve a complete spatio-temporal (2-D+T) decomposition. Motion-compensated filtering leads to multiresolution applications. It leads to optimum and adaptive signal-to-noise decomposition procedures based on the temporal correlative content. Such properties allow enhancing tasks like temporal interpolation, image sequence smoothing, and restoration. Simulation results are presented in this paper to illustrate the field of image sequence coding.

Progressive image transmission (PIT) techniques using the multiresolution wavelet decomposition are extended to other wavelet packet bases. Theses techniques are based on progressive selection and progressive quantization of subbands, with fixed or varying wavelet decompositions for the successive stages of transmission. The coding parameters of these stages are optimized under rate constraints, in accordance with two different approaches: a step by step optimization, and a global optimization for the whole transmission. Extending PIT techniques from multiresolution decomposition to other wavelet packet bases brings improvements in the successive qualities of reconstruction of an image.

An irregular adaptive sampling algorithm for a compact image representation is presented. A scattered data interpolation algorithm based on an inverse distance weighted method has been chosen. The goal is to compute the values and the locations of a fixed number of samples to approximate efficiently a natural grey-level image. These values are found by the means of a relaxation process. In the context of image coding, the sample values and sample locations have to be coded. A quantization is applied on the sample values and thus reduces the amount of information. The samples location are represented through a binary image which is coded using an arithmetic coder.

In this paper we address an important issue in motion analysis: the detection of moving objects. A statistical approach is adopted in order to formulate the problem. The inter-frame difference is modeled by a mixture of Laplacian distributions, and a Gibbs random field is used for describing the label set. A new method to determine the regularization parameter is proposed, based on a voting technique. Then two different multiscale algorithms are evaluated, and the labeling problem is solved using either ICM (iterated conditional modes) or HCF (highest confidence first) algorithms. Experimental results are provided using synthetic and real video sequences.

Motion estimation is crucial in domains such as robot vision or image sequence coding. Fast or real time computing is often required but is problematic due to computational complexity. In this paper, an algorithm using multiresolution multigrid Markov random fields is parallelized. Such realizations have already been studied for fine parallelizations on massively parallel machines. The goal of this paper is to achieve a coarse parallelization on a network of workstations, using the Parallel Virtual Machine (PVM) software package.

This paper presents a new temporal interpolation algorithm based on segmentation of images into polygonal regions undergoing affine motion. The goal of this work is to improve upon the block-based interpolation used in MPEG (B-frames). We briefly describe the region-based framework and the temporal linking algorithm that jointly provide the segmentation and motion parameters. We also present various applications of the proposed algorithm to temporal interpolative prediction. We examine one of these schemes in detail, including the special processing of occlusion areas. Results are illustrated by predicted images and using the MSE criterion we compare their quality with other schemes.

The paper presents new approaches in the field of motion-compensated spatio-temporal filters applied to digital image sequences. In time-varying imagery, the temporal correlation information of pixel intensities is folded by motions which may originate from both camera and object displacements. Motion-compensated filters are defined as temporal filters applied along assumed motion trajectories. As a matter of fact, the paper deals with three-dimensional spatio-temporal filters and aims at generalizing the motion-compensated temporal filtering process as the product of two distinct operators. The first operator depends only on the estimated motion parameters derived from both motion-based image segmentations and parametric affine modelings of regions in motion. The second operator analyzes only the correlations of image-by-image intensities measured along the assumed motion trajectories. Multiresolution filters or wavelets may be consequently applied along the motion trajectories to produce optimum and adaptive resulting procedures for purposes like spatio-temporal prediction, interpolation and smoothing. In the paper, applications are provided to cover the field of image sequence coding and interpolation.

The purpose of this paper is to introduce a new vector quantizer (VQ) for the compression of digital image sequences. Our approach unifies both efficient coding methods: a fast lattice encoding and an unbalanced tree-structured codebook design according to a distortion vs. rate tradeoff. This tree-structured lattices VQ (TSLVQ) is based on the hierarchical packing of embedded truncated lattices. So we investigate the design of the hierarchical set of truncated lattice structures which can be optimally embedded. We present the simple quantization procedure and describe the corresponding tree-structured codebook. Finally two unbalanced tree-structured codebook design algorithms based an the BFOS distortion vs. rate criterion are used.

Compression of segmented images is an essential step in segmentation-based approaches for image and video coding. This paper presents a lossless morphological sampling scheme for this purpose. With this sampling scheme, an image can be exactly reconstructed from its samples if it has been filtered by a morphological filter, but removal of any one sample would violate the exact reconstruction. Based on the sampling scheme, a new technique for segmented image compression is then developed. This technique can easily control coding errors and bit rates in the coding or decoding procedure.

In this paper we present an approach of selective compression of an image sequence based on a priori selection of region(s) of interest (ROI). This method relies on a given motion-based segmentation analysis. The problem of selective compression based on the concept of inhomogeneous spatial reconstruction quality is considered in the context of hybrid DPCM subband coding. Both spatial and frequency localization of the subband representation are explicitly used. Hierarchical compression is applied through adaptive quantization in the spatio-frequency domain. A weighted distortion metric is used to introduce both a priori and velocity-based visual masking.>

We have investigated an original motion estimation method that exploits several frequency subbands of wavelet pyramids using a multigrid-multiconstraint strategy. A recursive and iterative solution based on the Wiener approach allows to take into account the reconstruction quality criterion that is crucial to image coding. Experiments show its performances in terms of compression ratio, reconstruction quality, and reduction of implementation complexity compared to a monoresolution case.

An polynomial transform is used to decompose frames of an image sequence. For this purpose, a multiresolution expansion of images is constructed. The coefficients of the expansion form a pyramidal or hierarchical polynomial transform covering the frequency domain in octave bands. How image sequences are coded by a pel-recursive modified Wiener based motion estimation algorithm, applied to the multiresolution representation using a multigrid-multiconstraint scheme, is described. Simulation results show sequences reconstructed with broadcast quality.>

Motion analysis studies provide efficient estimation schemes for estimating dense optical flow fields. However, for image sequence coding where motion features have to be transmitted, an adaptive compact motion representation is required to extract temporal redundancy from an image sequence with a minimum amount of side information and, nevertheless, with a high quality of reconstruction. The advantages of region-based motion descriptions are illustrated and compared to those of the usual local ones (i.e., only two purely translational components of apparent motion). A new relaxation method for compact quantitative motion estimation using a hierarchy of motion models is presented.>

A new method is proposed to estimate global motion parameters from an unusual initial dense velocity field. The authors first want to obtain a compact representation of a dense velocity field and compute a spatiotemporal motion-based segmentation; then these parameters are used as initial values of a cost-function minimization algorithm and applied within a motion compensation loop for image sequence coding. Promising results are obtained on real TV image sequences. A compact motion representation is generated at each frame, and a quite interpretable qualitative and quantitative motion field is synthesized. Moreover, high quality of reconstruction and motion interpretation is obtained using the minimization stage.>

The authors present motion estimation methods applied to 3DTV image sequence coding schemes using motion compensation predictions. For 3DTV image sequence analysis and transmission, it is possible to jointly estimate in a coherent fashion two apparent motion vector fields subject to stereoscopic constraints. Locally, initial image data are grouped within two views (the left and right ones) at two successive time samples and spatiotemporal coherence has to be used to enhance motion vector field estimation. Several levels of coherence have been tested. The most promising approach, which consists in a joined stereoscopic motion estimation subject to motion and disparity constraints, is emphasized. Finally, results are shown on real stereoscopic 3D TV image sequences.>

Implementation of a set of pyramid transforms in a hierarchical approach for a real television sequence which consists of a 3-D data information set is presented. The investigation is based on several visual and objective comparative criteria, including the relevance to the motion information. Several pyramid transforms for image sequence coding are investigated and compared. The nonorthogonal pyramid transforms are separated from the orthogonal transforms. The nonorthogonal pyramid transforms are easily implemented, permit exact reconstruction, and accept the introduction of an error-compensation method if the different levels are coded but the nonorthogonality of the decomposition causes an increase in the number of transform samples and a high global entropy value (a disadvantage for data compression). The wavelet pyramid transform is more computationally complex but gives the best results for global entropy values. The quincunx pyramid seems to be the most interesting, since it features many advantages of the other including the visual representation of motion information through a pyramid sequence.>

A fixed-rate image-sequence coding method has been designed for the transmission of television signals of broadcast quality over a 34 Mb/s channel. The codebook replenishment algorithm for classified pruned tree-structure vector quantization in the image-sequence coding method is discussed. Each field of the image sequence is processed as follows: it is first segmented into 4*4 blocks, which are classified and then cosine transformed; visual thresholds are applied to the resulting coefficients while the mean is quantized, coded, and sent separately. Quantization follows, using pruned tree-structured vector quantizers. The codebook replenishment algorithm then modifies the codebook trees according to the properties of the current field.>

This paper describes a new method which combines vector quantization (VQ) and orthogonal transform coding schemes. The use of these two techniques in image coding allows efficient bit-rate reduction (about 8-10) with an original-like reconstruction quality and easy introduction of visual properties both in classification procedure and codebook design. Recent work, including classification algorithm and transformed coefficients coding is presented. Then, image sequence coding using VQ is discussed and some strategies for codebook refreshment are proposed.

We present in this paper an algorithm for motion estimation pel by pel in a sequence of television pictures. This estimator which will be used for a motion compensated coding, must be fitted any type of scene or motion. Techniques we used, are adaptative algorithm with detectors of ruptures without delays. Hence we estimate an a priori motion information on spatial edges where are chiefly located the abrupt changes in motion.