Catálogo de publicaciones - libros

H.264/AVC is the newest international video coding standard. This paper presents a novel hardware design for CABAC decoding in H.264/AVC. CABAC is the key innovative technology, but it brings huge challenge for high throughput implementation. The current bin decoding depends on the previous bin, which results in the long latency and limits the system performance. In this paper, the data hazards are analyzed and resolved using the algorithmic features. We present a new pipeline-based architecture using the standard look-ahead technique where the arithmetic decoding engine works in parallel with the context maintainer. An efficient finite state machine is developed to match the requirement of the pipeline controlling and the critical path is optimized for the timing. The proposed implementation can generate one bin per clock cycle at the 160-MHz working frequency.

Streaming media has contributed to a significant amount of today’s Internet Traffic. One solution of to solve this problems is using streaming proxy. There are two categories in streaming proxy; that is for homogeneous and heterogeneous client. The transcoding proxy can be used for heterogeneous client. The traditional proxy considers only a single version of the objects, whether they are to be cached or not. However the transcoding proxy has to evaluate the aggregate effect from caching multiple versions of the same object to determine an optimal set of cache objects. And recent researches about multimedia caching frequently store initial parts of videos on the proxy to reduce playback latency and archive better performance. Also lots of researches manage the contents with segments for efficient storage management. In this paper, we propose the efficient proxy policy that combines the segment-based caching mechanism and aggregate effect at transcoding proxy. The results demonstrate that the proposed algorithm outperforms in delay time, byte-hit ratio and the amount of transcoding data than other methods.

Implementing reliable Video on Demand (VoD) systems over the Internet, which is inherently best-effort, is a challenge. Distributed streaming for Video on Demand addresses this challenge with a combination of two techniques. The first, Distributed Video Streaming using Multicast (DVSM) involves video streaming from multiple servers to overcome path congestion by exploiting path diversity. The second technique, Asynchronous Hybrid mechanism for Video on Demand, implements a segmentation-based periodic broadcast to effectively utilize network bandwidth and decrease latency. The combination involves devising new algorithms for bandwidth estimation, segment partitioning and scheduling. A simulation of our proposed solution demonstrates its effectiveness. Specifically the results show, the prompt reaction of our strategy to congestion, and, the effect the various parameters have on system performance. The results shed light on parameters that can be fine-tuned for an effective VoD system.

A Body Area Network (BAN) is a body worn system which provides the user with a set of mobile services. A BAN incorporates a set of devices (eg. mp3 player, video camera, speakers, microphone, head-up display, positioning device, sensors, actuators). A BAN service platform for mobile healthcare and several health BANs targetting different clinical applications have been developed at the University of Twente. Each specialization of the BAN is equipped with a certain set of devices and associated application components, as appropriate to the clinical application. Different kinds of clinical data may be captured, transmitted and displayed, including text, numeric values, images and multiple biosignal streams. Timely processing and transmission of such multimedia clinical data in a distributed mobile environment requires smart strategies. Here we present one approach to designing smart distributed applications to deal with multimedia BAN data; namely the context awareness approach developed in the FREEBAND AWARENESS project.

Intra video coding is a common way to process video material for applications like professional video editing systems, digital cinema, video surveillance applications, multispectral satellite imaging, HQ video delivery, etc. Most practical intra coding systems employ JPEG encoders due to their simplicity, low coding delay and low memory requirements. JPEG2000 is the main candidate to replace JPEG in this kind of applications due to the excellent R/D performance and high coding flexibility. However, its complexity and computational resources required for proper operation could be a limitation for certain applications. In this work, we propose an intra video codec, M-LTW, which is able to reach very good R/D performance results, as well as JPEG2000 or H.264 INTRA, with faster processing and lower memory usage.

In this paper, we propose a parallel algorithm for H.264/AVC intra frame encoding by using the graphics processing unit (GPU). The proposed algorithm can handle 4x4 intra block prediction and reconstruction. By rearranging the encoding order of 4x4 blocks and modifying the architecture of H.264/AVC encoder, thirty times speed up can be achieved which utilizing the computing power of GPU without any loss in coding efficiency.

In this work, we investigate error patterns in compressed multi-view video signals and propose three error concealment algorithms, which can hide the effects of transmission errors efficiently. The proposed algorithms conceal a lost block by choosing and combining the best candidate blocks in the temporally adjacent frames or the inter-view frames at the same time instance. Simulation results demonstrate that the proposed algorithms effectively protect the quality of reconstructed videos against transmission errors.

In this paper, we address the problem of quality control for distributed video coding (DVC). In a pure DVC framework, the compression performance conflicts with the consistency of the video quality. A hybrid framework is proposed to solve this problem. Wyner-Ziv video coding and zero vector motion compensation are combined to keep the visual quality in consistency while maintaining satisfied compression efficiency. Simulation results show that the proposed framework can effectively reduce the variance of the video quality, and the compression performance is 1-5dB better than pure DVC framework worked in constant quality mode. We also evaluate the decoding efficiency of two multistage decoding (MSD) strategies in Wyner-Ziv coding. Simulate results show that passing soft decision from low level to high level works better in DVC.

In many rate control models, a uniform weighted distortion has been assumed; that is, the DCT coefficients of the motion-compensation difference frames (residues) conform to a uniform distribution. However, the residue after transform does not conform to a uniform distribution but approximate to a Laplacian distribution. In this paper, we first deduce a new distortion model with the assumption of Laplacian distribution of quantized DCT coefficients (called Laplacian distortion model, LDM), and then a more accurate rate-distortion model is proposed based on LDM. Experimental results on H.264/AVC show that our proposed method can improve PSNR up to 0.8dB compared to that of traditional TMN8; meanwhile, the mismatch of target bit rate and actual bit rate generated for coding can be controlled below 2.5%.

Previous work has shown that human faces under variable lighting conditions can be modeled by low-dimensional subspaces called illumination subspaces that can be computed using images under a universal lighting configuration. This configuration can be estimated using Harmonic images. However, harmonic images can only be obtained by using 3D information, and thus can be restrictive. In this paper, we overcome this limitation by presenting a completely data-driven method to find good universal lighting configurations. Motivated by the fact that affinity propagation clustering finds the cluster centers from the real images, we use affinity propagation clustering on real images taken under variable lighting conditions to find the cluster centres and use them to determine the lighting configuration. The illumination subspace for each individual is spanned by their images acquired in this lighting configuration. Matching is performed by comparing the distances to these individual illumination subspaces. Further, kernel methods are used to explore the non-linear structures of the illumination cone and carry out the illumination subspace methods in the kernel induced feature space. Experiments conducted on the Extended Yale Face B database demonstrate that the configuration obtained by our method is better than earlier recommended configurations. We also demonstrate that our technique is robust to pose variations using the CMU PIE database.