Image Enhancement Under Low Light Conditions

Pictures taken at night, cloudy, fog and haze have low contrast, low signal to noise ratio which restricted it applied in subsequent image feature extraction, image analysis and recognition. So it is vitally important and practical to enhance the low lighting video.This thesis introduced image enhancement algorithm based on the dark channel prior. For different illumination image and fog degraded image, we analyzed the characteristics of the atmosphere light intensity under the illumination, simplified atmospheric scattering model. Then, we proposed image enhancement algorithm based on depth information. According to the characteristics of the algorithm, we used improved dark channel prior to calculate transmittance, and optimized the transmittance. Experimental results show that our algorithm eliminated the halo artifacts, kept more of the original image detail to achieve good image enhancement.This thesis also analyzed the atmospheric scattering model and Retinex theory. Based on "Retinex algorithm on changing scales for haze removal with depth map", we proposed image enhancement algorithm based on transmittance calculated by Retinex. The algorithm regarded reflected light estimated by Frankle-McCann Retinex algorithm as transmittance. To solve color deviation phenomenon occurred during enhanced the image, we proposed algorithm to maintain color. The algorithm can enhance the image while maintaining good color original image, made a very good visual effects.