Research on Fractal Image Compression using Least Absolute Deviation Approach

• Main Authors: Nuan-ying Chiu, 邱煖媖
• Other Authors: Yih-lon Lin
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/19999420865562995969

碩士 === 義守大學 === 資訊工程學系碩士班 === 97 === In this thesis, fractal image compression using least absolute derivation (LAD) approach is proposed. LAD regression has been successfully applied in many branches of science and engineering, mainly due to the robust characteristics, for especially impulse noise. In traditional FIC, we see that the salt and pepper noise may destroy the self-similarity property of natural images, and therefore the similarity in terms of MSE by standard FIC is actually not a good measure of similarity. This motivates us to introduce LAD algorithm to compute contrast scaling and brightness offset computation. In the LAD approach, instead of the mean squared error, the LAD is used to provide another similarity measure of image blocks. For practical implementation, the weighted median computation technique has been used to find the numerical solutions of the corresponding L1-norm regression problems. Simulation results have shown that LAD has good robustness against the outliers caused by salt and pepper noise, but it does not show significant improvement in image quality for bell-shaped noises such as Gaussian and Laplace noises.