The Study of Adaptiave KLT in Image Compression

• Main Authors: Chen, Ding-Chaing, 陳定強
• Other Authors: Chang Lena
• Format: Others
• Language: zh-TW
• Published: 1998
• Online Access: http://ndltd.ncl.edu.tw/handle/97234847014804250635

碩士 === 國立海洋大學 === 航運技術研究所 === 86 === Image compression techniques are mainly relied on analyzing the spatial or temporal correlation and exploiting the redundancy of an image to achieve lower bit rates for transmission and maintain the range of fidelity for the reconstructed data. While spatial correlation is readily exploited by various image compression techniques, relatively little attention has been given to spectral correlation across bands. In the thesis, we first introduce the KLT-JPEG method which incorporates the best methods available to fully exploit the spectral and spatial correlation in the data. KLT is theoretically the optimum method to spectrally decorrelate the data and standard DCT based JPEG image compression algorithm is considered to be the most viable practical technique available today. Then, we introduce the Fixed-Region KLT(FR-KLT) method. By this method, the image is partitioned into sub- regions with fixed size and KLT is applied to each sub-region. The simulation results shows that the compression ratio of FR- KLT is better than that of KLT-JPEG. Due to the region partition of FR-KLT cannot be adjusted according to the spectral characteristics of each sub-region, we propose a Variable-Region KLT(VR-KLT) method. Using this method, we can combine the sub- blocks with similar spectral characteristics into the same region and we obtain better performance in compression ratio than FR-KLT. However, VR-KLT could not achieve the purpose of fast compressing image data by calculating the eigen components of each sub-block at first. To overcome this problem, we develop the Adaptive Variable-Region KLT(AVR-KLT) method which incorporates an adaptive KLT algorithm and the VR-KLT method. The AVR-KLT method has a better performance in compression ratio and maintain the range of fidelity for the reconstructed data. Moreover, the AVR-KLT could accelerate the speed of compressing image data by implementing its parallel architecture in hardware. In our studies, we also test the AVR- KLT method for remote sensing satellite images, and the simulation results show that we could get high compression ratio for remote sensing satellite images.