International Journal of Research in Engineering and Science (IJRES) ISSN (Online): 2320-9364, ISSN (Print): 2320-9356 www.ijres.org Volume 6 Issue 1 ǁ 1. 2018 ǁ PP. 21-25 www.ijres.org 21 | Page Vision Image Enhancement Using Optimized Retinex Algorithm Gilbert Owusu-Amonoo Gyamfi 1 ,Dzagbletey Philip Ayiku 1 , *Heejune Ahn 1 1 Departmentof Electrical & Information Engineering, Seoul National University of Science and Technology, Republic of Korea Corresponding Author:Heejune Ahn ABSTRACT: The Retinex algorithm has been used for illumination compensation, especially for night vision image enhancement. The Multi-scale Retinex with Color Restoration Algorithm (MSRCR) shows very good performance but has a high time complexity as well as introduces the so-called „halo effect‟ in the enhanced images. In this paper, we extend the MSRCR algorithm to include threshold detection and Gamma correction for the enhancement of both night-time and daytime images while maintaining the quality of the output of the algorithm. The simulation results with night and day time images shows a reduction of the halo effect in the enhanced images while the overall quality of the image is maintained. Keywords: gamma correction, image enhancement, night vision image, optimization, Retinex algorithm --------------------------------------------------------------------------------------------------------------------------------------- Date of Submission: 26-12-2017 Date of acceptance: 12-1-2018 --------------------------------------------------------------------------------------------------------------------------------------- . I. INTRODUCTION The Nowadays cameras are being used for a lot of applications during the night with artificial lighting providing illumination for these cameras. However, there are some applications such as driving in which cameras do need to work in different lighting conditions, even when the illumination is very low. And with the rate of accidents occurring during the night increasing, vehicles are now using auxiliary driving systems of which image processing technology is a key component [1]. Some of these driving aids, such as rear backup cameras are quite important in preventing accidents, but they cannot be of much benefit in places with poor illumination. Most modern vehicles also do not have high-end camera systems which can aid night vision enhancement like a professional camera does.Therefore, the images have to be enhanced in order to enable the driver to see obstacles and other things. There are a lot of algorithms that can be used for image enhancement, such as Histogram Equalization, Adaptive Histogram Equalization (ALE), Contrast Limited Adaptive Histogram Equalization (CLAHE) and Retinex algorithms [2][3]. However, the Retinex algorithm or variants of it is one of the most widely used algorithms currently. The Retinex algorithm was derived from the Retinex theory of human colour constancy developed by Edwin Land in 1976 [4]. This theory basically explains how the human visual system contributes towards colour perception and how it can distinguish colours under varying levels of illumination. Based on this theory, numerous algorithms have been developed in order to enhance images taken in varying illumination conditions. Retinex algorithms model any given image S as a pixel-wise multiplication of two images: the reflection image R, and illumination image L, i.e. S = L * R, and estimate the reflection image from the varying illumination condition L.Some work has been done so far in order to use the Retinex algorithm to process and enhance night-time images to improve visual detail [1][3][5][6][11]. Kyung et al [7] worked on using the multi-scale Retinex (MSR) algorithm for real time processing of night scene images to enhance the visibility of images from the vehicle‟s camera. Jobson et al [8] proposed a variant of multiscale Retinex in order to fill the gap between colour images and the human observation of scenes. Li He et al [9] also came up with an enhancement algorithm based on the Retinex theory in order to enhance images that are weakly illuminated. Hanumantharaju et al [10] developed a Multi Scale Retinex Algorithm with a modified colour restoration technique in order to enable true colour constancy to be obtained as well as make the algorithm more streamlined.Although the Retinex algorithm is one of the most widely used algorithms, it does have some drawbacks. Most notable of these effects is the so-called “halo effect”, which gives images enhanced using Retinex a glow-like appearance. Other effects include the loss of visual detail and high computational complexity. Work has been done in order to reduce the effects of the basic Retinex algorithm such as the „halo- effect‟ and high time complexity [4]. This paper therefore proposes a modified version of the Retinex algorithm based on the work done in [8] and [10] to improve the quality of enhanced images, reduce the halo effect on the enhanced images as well as widen the scope of the algorithm in order to improve both daytime images and night-time images. This paper is organized as follows: Section 2 gives a brief review of the Multiscale Retinex Algorithm. Section 3 describes the improved Multiscale Retinex Algorithm with Colour Restoration. Section 4 provides the results of the simulation and the conclusion is provided in Section 5.