Research On Digital Video Monitoring Real-time Noise Reduction

Video Monitoring, because of its advantages of intuition, convenience, reliability,rich information, and so on, is widely applied in areas such as finance, commerce,traffic, house, and community, and plays an important role in monitoring andprotection of these areas. However, the existence of the noise makes it an inevitableproblem in real-time video monitoring, and an important factor to evaluate the qualityof video vision.Among the video denoising algorithms, the time domain filtering algorithm issuperior than spatial filtering algorithm in the protection of frame edge and improvingPSNR. However, the motion evaluation is necessary to take advantage of correlationof video frames in time domain, although, due to the existence of the noise, theaccuracy of the motion evaluation is influenced, the performance of the algorithm isinfluenced and then. To improve the accuracy and achieve compatibility amongnoises of many levels, an adaptive algorithm is essential, on which the research isimperative. At present, because of the high computation load, the application of thede-noising algorithms in real-time video monitoring is limited. For that, a time-savingvideo denoising algorithms for noises of many levels is eagerly needed. Besides, thespatial filtering algorithm and the threshold setting of motion detection is more or lessrely on prior knowledge of the noises, however, it’s a pity that none of the exist noiseestimation algorithms can achieve perfect results in high quality, high noise, largeamount of texture information and severely motion conditions. So, a steady noiseestimation algorithm with a higher accuracy is essential.In this paper, three efforts are involved:1. A differential video noise estimation algorithm based on block neighborhoodcorrelation is demonstrated, which takes full advantages of the video signalcorrelation in time domain, can achieve results with higher accuracy in highquality, high noise, large amount of texture information and severely motionconditions. 2. A threshold setting algorithm for adaptive motion detection is demonstratedwhich improves the algorithm of video real-time denoising based on spatio-temporal combination, and leads to a higher accuracy and an adaptive performance.3. A fast motion estimation algorithm based on block structure similarity is demonstrated, which combines image structure similarity theory. With this algorithm, the time saving desire is realized, a higher accuracy is achieved,and the real-time demand is satisfied.