DWT/NLM Based-Satellite Image Resolution Enhancement Using Multi Wavelet-Domain Technique

Abstract

ABSTRACT

Multi wavelet-based resolution enhancement models (REM) enhance the image resolution by guessing the high-frequency band information. These models have one of the biggest drawback is losing high frequency contents which results in artifacts and blurring. Discrete wavelet transforms (DWT) based REM generates artifacts and block-ness due to a DWT shift-variant property. In this paper, an inter sub-band correlation based on dual-tree complex wavelet transform (DT-CWT) and nonlocal means (NLM) is proposed. A satellite input image is decomposed by DT-CWT which is nearly shifting-invariant to obtain high frequency sub bands. Interpolation techniques  have been widely used for RE based on nearest neighbors. To interpolate the high frequency sub bands and the low-resolution (LR) input images, an ideal interpolation filter is used to analyze the correlation between sub bands in the higher level. Meanwhile, the high frequency sub bands are passed through an NLM filter to equip for the artifacts generated by DT-CWT. Combining the filtered high frequency sub bands and the LR input image using inverse DT-CWT is to ensure high resolution enhanced image. The experimental analysis shows superiority of our new proposed technique over the conventional and state-of-the-art RE techniques.

Keywords: Discrete wavelet transform, dual-tree complex wavelet transform, resolution enhancement, shift variant, image resolution, adaptive histogram equalization.