Showing papers by "Lai-Man Po published in 1996"

A novel four-step search algorithm for fast block motion estimation

Abstract: Based on the real world image sequence's characteristic of center-biased motion vector distribution, a new four-step search (4SS) algorithm with center-biased checking point pattern for fast block motion estimation is proposed in this paper. A halfway-stop technique is employed in the new algorithm with searching steps of 2 to 4 and the total number of checking points is varied from 17 to 27. Simulation results show that the proposed 4SS performs better than the well-known three-step search and has similar performance to the new three-step search (N3SS) in terms of motion compensation errors. In addition, the 4SS also reduces the worst-case computational requirement from 33 to 27 search points and the average computational requirement from 21 to 19 search points, as compared with N3SS.

A new center-biased orthogonal search algorithm for fast block motion estimation

Abstract: A new center-biased orthogonal search algorithm (CBOSA) for fast block-based motion estimation is proposed. The new algorithm uses a center-biased search point pattern, which is based on center-biased motion vector distribution characteristic of the real-world image sequences, for improving the performance of the conventional orthogonal search algorithm (OSA). Simulation results show that the proposed CBOSA performs better than the OSA and the well-known three-step search in terms of compensation errors and search points requirement. In addition, the CBOSA also reduces the worst case search points requirement of the center-biased search algorithms from 33 of the new three-step search and 27 of the four-step search to 15 search points.

Partial distortion sensitive competitive learning algorithm for optimal codebook design

Abstract: An effective competitive learning algorithm based on the partial distortion theorem is proposed for optimal codebook design. Compared with some representative learning algorithms for codebook design, the proposed algorithm has consistently shown the best performance for designing codebooks of different sizes, especially large size codebooks.

Speeding up fractal image encoding by wavelet-based block classification

Abstract: The authors propose a novel fast fractal image encoding algorithm using wavelet-based block classification. The range blocks and corresponding domain blocks are classified into edge selective categories by the energy compacted wavelet coefficients. The searching is carried out between the same classes for range domain comparisons which significantly reduces the computational complexity. Experimental results show that the proposed method can speed up the fractal image encoding process by up to 12.3 times over the conventional full search method.

A Hybrid Adaptive Search Algorithm for Fast Block Motion Estimation

Abstract: This paper proposes a hybrid adaptive search algorithm (HASA) for block-based motion estimation. The proposed algorithm exploits the correlation between the block distortion measure (BDM) of the search origin (0,O) and its displacement from the motion vector to predict the range of motion. Based on the predicted motion type and the center-biased statistical distribution of motion vectors in low bit rate applications, one of the two block matching algorithms: Four-step search (4SS) or center-biased orthogonal search (CBOSA) is employed to find the motion vector. Experimental results show that HASA outperforms the well-known three-step search (3SS) and new three-step search (N3SS) in terms of number of search points while maintains acceptable mean square errors (MSEs). Moreover, it could effectively predict the stationary motion so as to stop the search at the first search point. The proposed algorithm is very suitable for low bit rate applications using software based video encoder.

Hierarchical Predictive Vq for Color Image Compression

Abstract: A hierarchical predictive VQ algorithm for color image compression is proposed in this paper. For encoding large size blocks, instead of using conventional bilinear interpolation, a new interpolation technique which make use of the neighboring reconstructed pixels and the lower right comer pixel is employed. The coding strategies for different block sizes are investigated, which involves the choice of color space and the use of combined or separate color components for vector quantization. Simulation results show that no significant blocking artifacts appear at the block boundaries. In addition, the encoding complexity for large size blocks can be maintained to an acceptable level.

Lossy Lzw Algorithm with Adaptive Threshold for Palettized Image Compression

Abstract: Color palettized image file formats especially the GIF format are widely used in the World Wide Web of the intemet. In these file formats. only lossless compression techniques are used for the encoding of palettized image. thus the compression ratio is only 1.2 to 2.5. To achieve higher compression ratio. lossy compression techniques have to be use. However. significant visual artifact due to lossy encoding of the palettized image may occur in the reconstructed images. and the compressed file format is not compatible to the conventional formats. To avoid these two problems. a Lossy LZW algorithm is proposed for palettized images compression. In the proposed algorithm, adaptive thresholds according to the human visual characteristics are used to consnain the color pixel’s error of the reconstructed image. In addition, the algorithm can be embedded in the GIF encoder so that the original GIF decoder can be used to decode the image compressed by the new algorithm.