Jae-Beom Lee

The VC-1 Video Coding Standard

VC-1 is a new video coding standard developed by Microsoft and standardized by the SMPTE. VC-1 is one of the three video compression algorithms standardized for high definition DVD. With high definition DVD players expecting to support MPEG-2, H.264, and VC-1, end users do not have to be concerned about the coding formats. The VC-1 standard offers a competitive quality-complexity tradeoff compared to H.264, especially for high-definition services. With a diverse digital video market, we can expect to see VC-1 co-existing with H.264 in the next generation of broadband and broadcast video services.

Probabilistic global motion estimation based on Laplacian two-bit plane matching for fast digital image stabilization

Digital image stabilization (DIS) is a technique to prevent images captured by a handheld camera from temporal fluctuation. This paper proposes a new DIS algorithm that reduces the computation time while preserving the accuracy of the algorithm. To reduce the computation time, an image is transformed by a Laplacian operation and then converted into two one-bit spaces, called and spaces. The computation time is reduced because only two-bits-per-pixel are used while the accuracy is maintained because the Laplacian operation preserves the edge information which can be efficiently used for the estimation of camera motion. Either two or four subimages in the corners of an image frame are selected according to the type of the image and five local motion vectors with their probabilities to be a global motion vector are derived for each subimage. The global motion vector is derived from these local motion vectors based on their probabilities. Experimental results show that the proposed algorithm achieves a similar or better accuracy than a conventional DIS algorithm using a local motion estimation based on a full-search scheme and MSE criterion while the complexity of the proposed algorithm is much less than the conventional algorithm.

An Efficient Algorithm for VC-1 to H.264 Video Transcoding in Progressive Compression

The high definition video adoption has been growing rapidly for the last two years. The two high definition DVD formats HD-DVD and Blueray have mandated MPEG-2, H.264 and VC-1 as video compression formats. The coexistence of these different video coding standards creates a need for transcoding. In this paper, an efficient transcoding algorithm from VC-1 video to H.264 video is discussed. While there has been recent work on MPEG-2 to H.264 transcoding, the published work on VC-1 to H.264 transcoding is non-existent. There is very limited amount of published work on VC-1. This paper gives a brief overview of VC-1 and discusses the opportunities for low-complexity tools for VC-1 to H.264 transcoding. The paper considers I and P frame transcoding from VC-1 to H.264 and proposes two approaches 1) low cost design and 2) high cost design

Programmable Deblocking Filter Architecture for a VC-1 Video Decoder

Although the current standards (MPEG1, MPEG2, MPEG4, H.261, H.263, and H.264/MPEG4 AVC) as well as the recent VC-1, present the same basic functional elements, i.e., prediction, transformation, quantization, and entropy encoding, important changes occur in the details of each functional element. One of the details concerns the elimination of the loss in interblock correlation due to block-based prediction, transformation, and quantization. In order to overcome the loss in blocking artifacts, a deblocking filtering method is necessary to maximize coding performance and consequently improve image quality. This letter describes a programmable VC-1 deblocking filter architecture with capabilities to support different standards. The architecture has been modeled, simulated, and implemented at the register transfer level. Results show a threefold performance improvement as compared to solutions where filtering algorithms are otherwise not hardwired. Results also point to parallelism based on existing data flow, and show that real-time requirements can be met.

Hardware/Software Partitioned Implementation of Real-time Object-oriented Camera for Arbitrary-shaped MPEG-4 Contents

Recently developed MPEG-4 part 2 compression standard provides a novel capability to handle arbitrary video objects. To support this capability, an efficient object segmentation technique is required. This paper proposes a real-time algorithm for foreground object segmentation in video sequences. The proposed algorithm consists of two steps: the first step that segments a frame into several sub-regions using spatio-temporal watershed transform and the second one that extracts a foreground object segment from the sub-regions generated in the first step. For real-time processing, the algorithm is partitioned into hardware and software parts so that computationally expensive parts are off-loaded from a processor and executed by hardware accelerators. Simulation results show that the proposed implementation can handle QCIF-size video at 15 fps to extract an accurate foreground object

A multi-standard micro-programmable deblocking filter architecture and its application to VC-1 video decoder

In order to overcome the loss in blocking artifacts due to block-based prediction, transformation, and quantization, a de-blocking filtering method is necessary to maximize coding performance and consequently improve image quality. This paper describes a programmable VC-1 de-blocking filter architecture with capabilities to support different standards. The architecture has been modeled, simulated and implemented in RTL. Results show a threefold performance improvement as compared to solutions where filtering algorithms are otherwise not hardwired. Results also point to parallelism based on existing data flow and show that real-time requirements can be met.

P-Frame Transcoding in VC-1 to H.264 Transcoders

VC-1 is now one of the three video coding standards for high definition DVD that include MPEG-2 and H.264. The coded is expected to be used in consumer electronic devices such as DVD and camcorders. The H.264 format has begun to see strong acceptance and is used in mobile devices such as iPod and mobile phones. While multi-format DVD players are expected to support the three high definition video coding formats, H.264 is expected to have broader support in devices and video download services. The need to move data among devices with different capabilities creates a need for transcoding. In this paper we present a P-frame transcoder for VC-1 to H.264 transcoding. The transcoder exploits the variable size transform used in VC-1 to select the variable block size for motion compensation in H.264. The transcoder reduces the complexity substantially without significant loss in quality.

Comparisons of fast 2D-DCT algorithms for parallel programmable digital signal processors

Recent advances in technology make it possible to integrate multiple processors into a single chip to build high performance parallel programmable digital signal processors (PPDSPs). These processors are expected to replace many dedicated digital signal processors to implement important image/signal processing algorithms such as discrete cosine transform (DCT). The paper addresses the issue of how to compare fast 2D-DCT algorithms when they are implemented on a PPDSP. Previously, the efficiency of these algorithms is compared based on the number of operations. This comparison is reasonable when these algorithms are implemented on a dedicated DSP. However, this comparison may not be suitable for general-purpose PPDSPs. The paper proposes to use three parameters, the number of data accesses, the number of communications, and the distance of communications, as new criterion for performance comparison of DCT algorithms. An algorithm-level technique is developed to estimate these parameters for DCT algorithms. The comparison results based on the parameters show that the algorithm proposed by Cho and Lee (1991) might be the best choice for a PPDSP unless it requires large overhead for communication between remote processors. In this case, the conventional row-column method with a fast 1D-DCT algorithm might be the most efficient.