A Retinex image enhancement based on L channel illumination estimation and gamma function
Yulei Huang, Yan Li, Yan Zhang
2018
Proceedings of the 2018 Joint International Advanced Engineering and Technology Research Conference (JIAET 2018)
unpublished
In order to improve the contrast of low illumination image, enhance the processing effect and eliminate artifacts, a color Retinex image enhancement algorithm based on L channel illumination estimation and two dimensional gamma functions is proposed. Firstly, the L image intensity is regarded as the product of correlation reflectivity and illumination through Lab color space transformation, and the intensity of global illumination is estimated by bilateral filtering. Then, two dimensional gamma
more »
... functions are used to complete the pixel mapping, and the local details can be adaptively adjusted. Experimental results show that compared with other Retinex image enhancement algorithms, the proposed algorithm enhances the effect more naturally and produces no enhancement artifacts, which can effectively preserve the local details of the image. Introduction In light of insufficient lighting conditions or performance limitations of photography equipment, there will be a large number of low-illumination images in daily life and work. At this time, there is an urgent need to enhance low-light images. In order to improve the visual effect of the image and increase the visual information of the image, the image needs to be enhanced. For color images, the amount of information they contain is more abundant than gray scale images, so the corresponding enhancement processing algorithm will be more complex [1] . Contrast enhancement algorithm is one of the hot topics in image processing. Most of the traditional contrast enhancement algorithms are widely used Retinex theory [1, 2] . Jobson et al. proposed a single-scale Retinex (SSR) method for image enhancement [3] . This method uses Gaussian filtering to evaluate and process the luminance of the input image. However, in the enhanced output image, halo artifacts often appear in local images with complex texture edge information, which is called color shift [4] [5] [6] . According to the principle of mimicking the operation of the human visual system, an enhanced algorithm based on Retinex can be used to enhance the output image with higher naturalness. Therefore, many researchers have recently proposed various enhancement algorithms based on Retinex [7-9]. The literature [10] uses bright pass filters and double logarithmic conversions to simultaneously process reflectivity and luminance, balancing the resulting image details and naturalness. However, none of these algorithms can meet the requirements of high naturalness and artifact-free phenomena. In order to maintain its original brightness and local details in enhanced low-light images, we need to do local adaptive brightness processing. Based on the above research and analysis, this paper presents a color Retinex image enhancement algorithm based on L-channel illumination estimation and two-dimensional gamma function. First, the algorithm obtains L-channel components through Lab color space transformation, and uses bilateral filtering to estimate the global illumination brightness. Then, pixel mapping is performed by using a two-dimensional gamma function to achieve local brightness adaptive adjustment. Finally, the enhanced global luminance component and adaptively adjusted local luminance are combined to generate the final enhanced image. Experimental results show that compared with other Retinex image enhancement algorithms, the proposed algorithm can maintain 312
doi:10.2991/jiaet-18.2018.55
fatcat:clkppuvwkbbqhdwqvlhsrvdvm4