Image enhancement algorithm combining multi-scale Retinex and bilateral filter

Under some special conditions, the quality of the inquired image is low, and the image has blurred characteristic, in order to improve the quality of the blurred image, this paper proposes a blurred image enhancement algorithm based on improved multi-scale Retinex. The shortcomings of the traditional methods are analyzed, by the idea of constraining weighted average to optimize traditional algorithm, and enhance detail information of blurred image, and multiple blurred image are utilized on the algorithm performance simulation test. The simulation results show that the blurred image enhancement algorithm of improved multi-scale Retinex, solves the problem of the traditional Retinex algorithm, speed up the running speed of the blurred image enhancement, makes the blurred image clearer, and obtain better visual effect. Introduction In the blurred conditions for image shooting, due to the impact of atmospheric scattering effects, the color and contrast of captured image degenerate, image quality is lower, adversely affect is produced to the outdoor video surveillance system, so the blurred image have to be processed, so as to improve the quality of the image, which has very important significance [1]. The main purpose of blurred image enhancement is to improve the image contrast and saturation, and keep chroma unchanged. The blurred image enhancement algorithm is divided into physical methods and image processing method currently [2]. The physical method requires the prior knowledge of sunny image at same scene, and some physical device, which is not convenient in practical applications [3]. Therefore, independence on the physical device, has become the main research direction of blurred image enhancement, the common methods include: histogram equalization, wavelet transform, homomorphic filtering and Retinex algorithm [4-7]. The histogram equalization method by increasing the detail contrast of the image, to improve the quality of the image, but the original image color is distorted; homomorphic filtering method can have better processing for the image of not uniform illumination, but has poor effect on blurred image [8]. The Retinex theory of color constancy proposed by Edwin Land, can enhance the contrast of the image, and provide a new enhancement method to solve the blurred image quality problems [9]. However, a large number of studies show that, the image output by Retinex theory have some defects, like halo and over enhancement, the complexity of the algorithm is high, and can lead to RGB color offset to gray [10]. Then, some scholars have put forward a series of Retinex image improved enhancement algorithm, to some extent improve the blurred image quality, the enhancement effect is superior to the traditional Retinex algorithm, but the algorithm complexity is much higher [11]. Zhao Quanyou etc. obtained better multi-scale enhancement effect with the nonlinear Sigmoid, but it can only enhance dark region, and also not suitable for blurred image enhancement [12]. The traditional Retinex image enhancement algorithm and defects In the traditional method, dark channel enhancement method is utilized for a scene image shoot under the fog environment, but the enhancement effect is not good. Therefore, this paper proposes International Conference on Automation, Mechanical Control and Computational Engineering (AMCCE 2015) © 2015. The authors Published by Atlantis Press 1221 blurred video image enhancement algorithm of improved multi-scale weighted average Retinex algorithm. The multi-scale weighted average Retinex algorithm is to determine a preferred enhanced block from many blocks need to be enhanced with regarding the edge pixels as the center, the image texture structure information in unit time is ( ) , ; G x y t , intuitionistic blurred sets of image texture subspace are defined as conduction function: 0 ( ) ( , ) ( , ) lim[ ] x u u u u x t p x t x x σ σ ∆ → − + ∆ ∂ = = − ∆ ∂ (1) Adopting multi-scale weighted average Retinex algorithm to process adaptive enhancement for gray pixels value of image, as: 2