A Study on Color Image Decimation for MPEG-4 Dynamic Resolution Conversion

• Main Authors: Yu-Chih Tsai, 蔡育誌
• Other Authors: Pi-Chi Chen
• Format: Others
• Language: zh-TW
• Published: 2004
• Online Access: http://ndltd.ncl.edu.tw/handle/02025231367871316482

碩士 === 義守大學 === 資訊管理學系碩士班 === 92 === High-speed digital data transmission services with mobile equipment are becoming available. Though the visual signal is one of the expected media to be used with such transmission capabilities, the bandwidth of visual signal is much broader than the provided transmission bandwidth in general. Therefore efficient video encoding algorithms have to be introduced. The ITU-T Recommendation H.263 and ISO/IEC MPEG-4 are very powerful encoding algorithms for a wide range of video sequences. But a large amount of bits are generated in highly active scenes to encode them using such conventional methods. This results in frame skipping and degradation of decoded picture quality. In order to keep these degradations as low as possible, MPEG-4 proposed a Dynamic Resolution Conversion (DRC) method of the prediction error. In the method, a reduced resolution encoding is carried out by using decimation when the input scene is highly active. There are lots of methods of decimation and interpolation which often be applied to Image Resolution Conversion and Image Change of Scale, and the best method is Cubic-Spline Interpolation (CSI). In this thesis, we proposed a fast Cubic Spline (CSP) method based on the properties of color image space in common use to improve CSI scheme that is not easy to utilize the Block-based decimation method. Also CSP scheme adapts easily to more different color image space than CSI scheme does. Besides, we applied CSP to MPEG-4 DRC, and experiment results show that the proposed scheme can improve both coding efficiency and picture quality in MPEG-4 DRC.