Evaluation of Wavelet Transform Families in Image Resolution Enhancement
Pooneh Bagheri Zadeh, Akbar Sheikh-Akbari · 2015
Wavelet transform have been used in many image enlargement techniques and shows superior results compared to conventional image resolution enhancement techniques.However, the choice of wavelets for image enlargement has not been fully investigated.This paper presents an investigation on the performance of a generic wavelet based image enlargement technique, with respect to the choice of wavelets.A Wavelet based image enlargement technique using cycle-spinning is employed for the purpose of this investigation.Different wavelet transforms from wavelet transform families such as Daubechies, symlet, Coiflet, Biorthogonal, Reverse Biorthogonal and Discrete Meyer wavelets are used in this investigation.A summary of the process for this investigation is as follows: A high resolution image is first generated from an input low resolution image, using an inverse wavelet transform, where the coefficients in high frequency subbands are sets to zero.The cycle spinning method is then used to remove the ringing artefacts of the resulting high-resolution image.The generated results indicate that the performance of the image enlargement technique is a function of the chosen wavelet and also the frequency components or texture of the image.Results implies that the enlargement method produces superior results using Daubechi 9/7 and Symlet 2 wavelets, when enlarging images containing significantly higher frequency components.The technique generates superior results when dealing with images containing lower high frequency components using Daubechi 9/7 and LeGall 5/3.