基于暗原色先验的低照度视频快速增强算法-信息与通信工程专业论文.docxVIP

II II ABSTRACT Multi-media technology and video monitor technology have already become the common communication and record tools. But when shooted in low illumination environment such as conference room, cinema and the dark, the video will be degenerated with low contrast, low signal-to-noise ratio and poor visual effect. Under this circumstance, it is difficult to show the details of images, which makes subsequent image feature extraction, image analysis and recognition restricted. So it is vitally important and practical to enhance and denoise the low lighting video. This kind of technology can be applied in monitoring and control system, pattern recognition, medical image analysis, etc. This paper introduces traditional low lighting enhancement and denoising theories, illustrates their different processes and points out the defects of above methods in solving the local details brightness, noise, real-time processing, etc. Compared to traditional enhancement algorithm, existing “low lighting video enhancement algorithm using dark channel prior” has higher fidelity with more details. But because of the underexposure of low lighting images, the noise in images after enhancement will badly degenerate images. Therefore, this paper proposes an improved efficient method based on the dark channel prior to enhance low lighting video. Based on existing “low lighting video enhancement algorithm using dark channel prior”, this paper takes advantage of both the information of time domain and space domain and bilateral filter to denoise the low lighting video. This algorithm can not only denoise images, but also remain more motion information. Meanwhile, the algorithm is optimized to adapt to GPU CUDA programming by partitioning and paralleling the data. By using GPU CUDA parallel programming, the algorithm can achieve real-time processing performance. Experimental results show that the algorithm which achieves the goal of gaining clearer and more natural high quality video ca