Wavelet based multilevel un-sharp masking using OpenCL

Abstract

A wavelet based multi-level adaptive unsharp masking technique for image sharpening is proposed. It does sharpening at multiple levels of DWT with a small and fixed size Gaussian kernel and automatically adjusts for different amount of blurring in different direction. The proposed method is free from kernel estimation. The algorithm is designed to process in a heterogeneous environment consisting of CPU and GPU using OpenCL for fast processing. Significant performance enhancement has been achieved as compared to the CPU implementation and can be implemented in mobile devices for near real-time performance.