论文信息 - A Parallel Architecture for Stereoscopic Processing

A Parallel Architecture for Stereoscopic Processing

An embedded pipeline/parallel architecture to support an extended quad-tree algorithm suitable for real-time estimation of the dense disparity map (DDM) for stereoscopic image processing is proposed. The system performance has been analyzed by several simulations to qualify the results by both an objective measurement (Mean Square Error) and a subjective assessment (output images). The proposed extended quad-tree is based on the block-matching algorithm, then a fine-grain granularity analysis to estimate the DDM leads us to a systolic array design for the basic Processor Element. This basic design has been utilized to the next levels quad-tree's Processor Elements design.

Bruno Ciciani | Milton Romero

[1] Liang-Gee Chen,et al. An efficient and simple VLSI tree architecture for motion estimation algorithms , 1993, IEEE Trans. Signal Process..

[2] Bruno Ciciani,et al. Depth-controlled 3D TV image coding , 1998, Electronic Imaging.

[3] Dimitrios Tzovaras,et al. Three-dimensional camera motion estimation and foreground/background separation for stereoscopic image sequences , 1997 .

[4] S. Kung,et al. VLSI Array processors , 1985, IEEE ASSP Magazine.

[5] Berthold K. P. Horn. Robot vision , 1986, MIT electrical engineering and computer science series.

[6] P. J. Burt,et al. The Pyramid as a Structure for Efficient Computation , 1984 .

[7] Bruno Ciciani,et al. A Fault-Robust SPMD Architecture for 3D-TV Image Processing , 1998 .

[8] P. Pirsch,et al. Advances in picture coding , 1985, Proceedings of the IEEE.

[9] Yu Hen Hu,et al. A novel modular systolic array architecture for full-search block matching motion estimation , 1995, IEEE Trans. Circuits Syst. Video Technol..

[10] Azriel Rosenfeld,et al. Multiresolution image processing and analysis , 1984 .