Tarek Ouni

New low complexity DCT based video compression method

Generally, video signal has high temporal redundancies due to the high correlation between successive frames. Actually, this redundancy has not been exploited enough by current video compression technics. In this paper, we present a new video compression approach which tends to hard exploit the pertinent temporal redundancy in the video frames to improve compression efficiency with minimum processing complexity. It consists on a 3D to 2D transformation of the video frames that allows exploring the temporal redundancy of the video using 2D transforms and avoiding the computationally demanding motion compensation step. This transformation turns the spatial-temporal correlation of the video into high spatial correlation. Indeed, this technique transforms each group of pictures to one picture eventually with high spatial correlation. Thus, the decorrelation of the resulting pictures by the DCT makes efficient energy compaction, and therefore produces a high video compression ratio. Many experimental tests had been conducted to prove the method efficiency especially in high bit rate and with slow motion video. The proposed method seems to be well suitable for video surveillance applications and for embedded video compression systems.

Lossless image compression using gradient based space filling curves (G-SFC)

In most classical lossless image compression schemes, images are scanned line by line, and so, only horizontal patterns are effectively compressed. The proposed approach attempts to better explore image correlation in different directions by adopting a context-based adaptive scanning process. The adopted scanning process aims to generate a compact one-dimensional image representation by using an image gradient based scan process. This process tries to find the best space-filling curve that ensures scanning the image according to the direction where minimal pixels’ intensity change is found. Such scan process would reduce high frequency data. It is used in order to provide an easily compressible smooth and highly correlated mono-dimensional signal. The suggested representation acts as a pre-processing which transforms the image source into some strongly correlated representation before applying coding algorithms. Based on this representation, a new lossless image compression method is designed. Our experimental results show that the proposed image representation is able to significantly improve the signal proprieties in terms of correlation and monotony and then compression performances. The suggested coding scheme shows a competitive compression results compared to conventional lossless coding schemes such as PNG and JPEG 2000.

Adapted Scan Based Lossless Image Compression

This paper deals with the application of lossless compression algorithms to two-dimensional curves scanned images. An image is scanned along a space filling curve (SFC) so as to exploit inherent coherence in the image. The used SFC is determined by a gradient based method allowing the detection of global pixel’s change direction for each image block. The resulting one-dimensional representation of the image has improved auto-correlation compared with universal scans such as the Peano-Hilbert space filling curve. Combined with conventional coding algorithms the proposed algorithm shows significant compression efficiency improvement. The new algorithm used for SFC determination is presented and used as an input to conventional coding schemes.

New non predictive wavelet based video coder: performances analysis

A non-predictive video coding is a new branch of emerging research area in video coding, where the motion estimation/compensation or prediction step in the temporal domain is omitted. One direction was to look for the exploitation of temporal decomposition of video frames. The proposed method consists on 3D to 2D transformation of the temporal frames that allows exploring the temporal redundancy of the video using 2D wavelet transforms and avoiding the computationally demanding motion compensation step. Although the many advantages presented by the proposed coder, some annoying artifacts still exist. In this paper, we will explore the performances of the proposed method and try to better show what it actually offers to users. The paper presents also the extensions chosen in order to reduce the perceived artifacts and increase the perceptual as well as objective (PSNR) decoded video quality, which is actually competitive with state-of-the-art video coder algorithms, especially when low computational demands of the proposed approach are taken into account.

Multi-shot person re-identification approach based key frame selection

This paper presents a novel approach to solve the problem of person re-identification in non-overlapping camera views. We propose an appearance based method for person re-identification that condenses a set of frames of the same individual into the multi-class classifier SVM (Support Vector Machine). Still, the choice of different and most expressive frames for each target is very challenging. Besides, efficient person re-identification algorithms are computationally expensive due to the big amount of data used. One of the originalities of our method is how to select different shots during person tracking within each camera to guaranty efficient person re-identification. We evaluate our approach on the publicly available PRID 2011 multi-shot re-identification dataset and demonstrate some performance in comparison with the elimination of the proposed key frames selection.

Mono-camera person tracking based on template matching and covariance descriptor

This article presents a simple and efficient approach to persons tracking within large scale environment. The proposed approach is a point matching tracking algorithm based on a covariance descriptor. Object tracking, in general, is a challenging problem. Difficulties in tracking objects can arise due to abrupt object motion, changing appearance patterns of the object and the scene and partial and total occlusions. Tracking is usually performed in the context of higher-level applications that require the location and appearance of the object in every frame. Typically, assumptions are made to constrain the tracking problem in the context of a particular application. The ultimate purpose of the proposed approach is to propose an efficient tracking algorithm as a way for real time multi-shot re-identification. This approach is evaluated using standard datasets.

Scan Based Lossless Image Compression Scheme

In this paper we propose a new loss less image compression method. The suggested method relies on novel selective scan process that aims to scan the image in the direction where minimal pixel intensity change is recorded. Such scan process would reduce high frequency data, in order to provide a smooth and high correlated mono-dimensional signal easy to compress. This scan process is followed by an adapted coding scheme based on DPCM and Huffman algorithms. The experimental results show a significant compression improvement of the suggested method as compared to two common loss less coders in the same class.

Images Selection and Best Descriptor Combination for Multi-shot Person Re-identification

To re-identify a person is to check if he/she has been already seen over a cameras network. Recently, re-identifying people over large public cameras networks has become a crucial task of great importance to ensure public security. The vision community has deeply studied this area of research. Most existing researches rely only on the spatial appearance information extracted from either one (single-shot) or multiple images (multi-shot) for each person. Actually, the real person re-identification framework is a multi-shot scenario. However, to efficiently model a person’s appearance and to select the most informative samples remain a challenging problem. In this work, an extensive comparison of descriptors of state of art associated to the proposed frame selection method is considered. Specifically, we evaluate the samples selection approach using different known descriptors. For fair comparisons, two standard datasets PRID 2011 and iLIDS-VID are used showing the effectiveness and advantages of the proposed method.

Optimized Parallel Model of Human Detection Based on the Multi-Scale Covariance Descriptor

Human detection based on the multi-scale covariance descriptor outperforms many other antecedent descriptors. However, it has the disadvantage of being highly time consuming. The complexity of this type of application intensifies the need of multiprocessor architecture (MPSOC) to meet real time constraints. A well-balanced application model is necessary for an efficient implementation into MPSOC architecture. In this paper, a high-level independent target-architecture parallelization approach is used to propose an optimized parallel model of a multi-scale covariance human detection system. The main characteristic of this approach is the exploration of both task and data levels of parallelism. For this end, an initial model is proposed. This model is implemented and validated at a high level interface. The potential bottlenecks of this model are identified using communication and computation workload analysis. Based on this analysis, an optimized parallel model with maximum workload balance is provided. Results reveal that the obtained parallel model has more than four times lower execution time in comparison with the sequential model.

New Accordion based video compression approach

In this paper, we propose a new approach based on the accordion transform that tends to exploit the both intra and inter frame correlation. The major novelty of the new approach compared to classic one is the dynamic strategy of video frames scan. The direction of the frames scan will depends on a gradient based algorithm which tries to predict the direction where minimal pixels intensities changes are recorded. Both objective (PSNR) and subjective quality evaluation prove the improvement of the compression efficiency, especially in fast and uniform motion video sequences which usually represents the weakness of the classical approach.