Research Of Night Color Image Enhancement Technology

Due to the uneven distribution of light at night, the quality of the night color image is usually poor such as low image contrast, low brightness and less texture. Poor image quality will make people can not clearly identified in the image scene, and the poor quality of the images often can not be directly used for analysis and processing. In public security, military reconnaissance, nondestructive testing and other areas, night color image enhancement technology has a bright future. In order to better identify the image scene, better use of poor-quality images, it is important for enhancing the image obtained at night, mainly to ensure that this enhanced image contrast and brightness of the image can be restored to normal level.Image enhancement techniques apply to the single nighttime image to improve the visual effect,making the original image which is not clear become clear or emphasizing certain features of interest, so that the image-processed is in line with the observed habit of the human eye and meet the requirements of the machine identification. Image enhancement techniques can be divided into two categories of unified approach to space and non-uniform method to space. Unified approach to space include: logarithmic compression, gamma correction, histogram equalization, linear stretching; non-uniform method to space include: local histogram equalization, the method based on contrast sensitivity of human eye,the method based on Retinex and the like. Through the study on the optimization method、the method based on Retinex and dictionary learning method, this paper proposes three new night color image enhancement algorithms. The main work and innovation of this paper is as follows:(1)Night color image enhancement via optimization of purpose and improved histogram equalization: First, enhance the contrast of the source image and reserve details furthest through improving the image gradient values using the method of optimization. Then, enhance the image by improved histogram equalization which increases the probability of pixel values of small probability. Finally, enhance the image brightness through gamma correction. Through a lot of experiments, the algorithm enables enhanced image of the scene appear more clear information, suppress "Halo artifact" phenomenon, and the results can meet people’s subjective feelings.(2) Night color image enhancement via statistical law and Retinex: The algorithm analyzes the transformation relationship between the nighttime image and illumination image by the algorithm of Michael Elad and MSRCR algorithm. Through this transformation, we can accurately and quickly get the illumination image. Then, we can get the resulting image successfully based on the retinex. Our algorithm can greatly enhance the image contrast and brightness, recover image details, eliminate the “halo effect” efficiently.(3) Night color image enhancement method based on dictionary learning and sparse representation: The algorithm first proceeds dictionary learning for each image block of the input image and the corresponding normal image by K-SVD method, gets two over-complete dictionaries which have same sparse representation for the input image and the corresponding enhanced image. Thus, we can obtain the sparse representation of the input image in the corresponding dictionary by the OMP algorithm, and then we can obtain the enhanced image obtained by the obtained sparse representation and the other dictionary. Experiments show that the algorithm can effectively enhance the brightness and contrast of the night color image, contrast, recover image details, eliminate the "halo effect", suppress noise. The results are in line with the human eye habits.