Quaternion Fourier transform based alpha-rooting method for color image measurement and enhancement

Abstract

The aim of this paper is to present a transform-based framework for color image enhancement which processes all three color components (R,G,B) simultaneously. In this framework, the image enhancement procedure is supported by a two-dimensional discrete quaternion Fourier transform and the new multi-frequency band α-rooting concept. Moreover, we introduce a simple measure to evaluate the quality of enhanced color images. Additionally, we demonstrate the performance of the proposed algorithms against recently proposed methods which separately process color channels. Preliminary results show that the presented algorithm out-performs other enhancement methods such as the traditional Fourier transform-based alpha-rooting and the multi-scale Retinex algorithms. The proposed algorithms are fast and simple in implementation and design which makes them very practical for image enhancement.