Proposal for a new method to speed up local color correction algorithms

Abstract

There is a class of non-linear filtering algorithms for digital color enhancement characterized by data driven local effect and high computational cost. In this paper we propose a new method, called LLL for Local Linear LUT, to speed-up these filters without loosing their local effect. Usually LUT based methods are global while our approach uses the principles of LUT transformation in a local way. The main idea of the proposed method is to apply the algorithm to a small sub-sampled version of the original image and to employ a modified Look Up Table technique to maintain the local filtering effect of the original algorithm. In this way three functions, one for each chromatic channel, are created for each pixel of the original image. These functions filter the original full size image in a very short time. We have tested LLL with two of these filters, a Brownian Retinex implementation (BR) and ACE (Automatic Color Equalization). The computational cost for this algorithms is very high. The proposed method increases the speed of color filtering algorithms reducing the number of pixel involved in the computation by sub-sampling the original image. Results, comparison and conclusion are presented.