Wavelet image compression by using hybrid kernel SVM

• Published in: 2008 International Conference on Machine Learning and Cybernetics Vol. 5; pp. 3056 - 3060
• Main Authors: Jia-Ming Chen, Lei Li, Ling-Ye Nie
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2008
• Subjects: Hybrid Kernel; Image coding; Image Compression; Kernel; Polynomials; Support vector machines; Transforms; VC-dimension; Wavelet domain; Wavelet Transform; Wavelet transforms
• ISBN: 1424420954; 9781424420957
• ISSN: 2160-133X
• DOI: 10.1109/ICMLC.2008.4620932

In this paper, we proposed a way through combining the support vector machines (SVM) with hybrid kernel and wavelet transform to compress the image. SVM regression could learn dependency from training data and realized compression by using fewer training point (support vectors) to represent the original data and eliminate the redundancy. Wavelet coefficients could be compressed based on this feature. Further more, the hybrid kernel applied can enhance the compress efficient and improve the picture quality by controlling the VC-dimension (Tan, 2004) of SVM. At last, we use the arithmetic coding to encode the dates from the output of the SVM and finish the image compression.