Light Random Sprays Retinex: Exploiting the Noisy Illumination Estimation

Abstract

In this letter, Light Random Sprays Retinex (LRSR), an improvement of the Random Sprays Retinex (RSR) algorithm is proposed. RSR is a white balancing algorithm for achieving local color constancy and image enhancement by using random sprays of the same size. The main problem of the original RSR is that the lower the number and size of the sprays, the greater the noise in the resulting image, which means that the number and size of sprays have to be relatively high in order to reduce the noise leading to a higher computation cost. The proposed improved algorithm is based on a new method to remove the noise in the resulting image thereby allowing only one spray of a smaller size to be used resulting in lower computation cost. By using interpolation the computation cost is reduced even further without a noticeable perceptual difference. The improvement is tested on a public database and is shown to outperform the original RSR in image quality and computation cost. The source code is available at http://www.fer.unizg.hr/ipg/resources/color_constancy/.

Improving the white patch method by subsampling

“…Estimating local illumination by using a relatively small pixel sample was shown to work well in the image enhancement Light Random Sprays Retinex algorithm [17] where there was no perceptual difference for images with illumination estimated for all pixels and with illumination estimated for 2.7% pixels with the rest being calculated using interpolation. Pixel illumination estimations were based on performing the white patch method on a relatively small number of neighborhood pixels selected with a random spray.…”

Color Rabbit: Guiding the distance of local maximums in illumination estimation

“…The Color Sparrow (CS) algorithm [15] is based on the Random Sprays (RSR) algorithm [16], which for every pixel centers N sprays of n randomly distributed pixels around it and recalculates intensities for all color channels as…”

“…Table I and Table II show the best means and medians of errors for the proposed method and for several other color constancy methods on several databases. The data was taken from [21], [25], [15], [8], and [22] or was calculated by using the available source code. On the ColorChecker and the GreyBall databases the proposed method outperforms all other low-level statistics-based methods and it compares well to the learning based methods.…”

“…The execution times that are given in Table III show that even though it is more accurate, the proposed method does not get behind with regard to execution time. All of the methods were executed with parameter values set for best accuracy and this values were taken from [15] and [25]. The code used in testing is provided at http://www.fer.unizg.hr/ipg/resources/color constancy/.…”

“…Another recent low-level statistics-based algorithm is the Color Sparrow (CS) algorithm [15], which is based on the Random Sprays Retinex (RSR) algorithm [16] and uses the same noise removal as the Light Random Sprays Retinex (LRSR) algorithm [17], the improved version of RSR. Both RSR and LRSR perform per-pixel illumination estimation and brightness adjustment and CS uses some of these local illumination estimations to combine them into a global one.…”