Minimization of boundary artifacts on scalable image compression using symmetric-extended wavelet transform

Abstract

Image compression techniques are routinely applied to conserve storage space and minimize bandwidth utilization in various video and communication applications. Wavelet transform is an efficient approach to reduce spatial redundancies without the annoying blocking artifacts at low bit rates. The underlying signal processing used in wavelet transform is the convolution between the decimated input signal and the wavelet filters. Since image signals are not continuous at the boundaries, problems of coefficient expansion and boundary distortion are faced in the implementation of the filtering on the finite length signal. Circular convolution instead of linear convolution can eliminate coefficient expansion but introduce boundary artifacts, especially when more levels of decompositions are involved to obtain scalable images. Symmetric-extended wavelet transform (SWT) introduced in this paper can greatly reduce boundary artifacts with improved performances in scalable image compression in comparison with the periodic-extended wavelet transform and the JPEG compression.