Application Research Of Video Interference Identification Based On Statistical Analysis

With the rapid development and gradual improvement of optical, information processingand multimedia communication technology, video information has been widely used in our dailyproduction and life. The era of big data in video industry has quietly arrived. However, due to theimpact of the installation environment and quantity of information acquisition equipment, theinformation processing algorithm, transmission bandwidth and storage costs, etc al. the needs tomonitoring, diagnosis and restoration of video quality has become a current priority. Interferenceidentification research is an important part of video quality monitoring, diagnosis and restoration,which is able to change the traditional way of artificial quality monitoring and the passive way ofsurveillance systems maintaining, to accomplish intelligent monitoring and diagnosis, to providea priori knowledge for restoration. Stripe interference, blocking interference and flickeringinterference in temporal domain is three common interference types affecting the quality ofvideo images. From the perspective of the statistical analysis, these three types of interferencehave been analyzed, and an effective no-reference identification solution has been presented.The movement in one direction of stripe interference in temporal domain and the translationof Fourier transform are utilized to stripe interference identification. Period of time, according tothe statistical characteristics of the peak in the Fourier magnitude spectrum can identify the stripeinterference.Identification of blocking interference takes advantage of a pseudo-periodic boundaryeffects of blockiness. Under the premise of a known block size and block boundary direction, theeffective blocking interference identification scheme has been proposed using the statisticalfeature of amplitude and phase spectrum in a certain frequency point.Analyzing the statistic properties of the differential signal of temporal flickeringinterference, it is found that the differential signal obey Laplace distribution. Under the premiseof Laplace distribution, using small probability event ideas and iterative fitting scheme, thedifferential signal of moving targets which impacts the measurement estimation of temporalflickering has been effectively eliminated. Meanwhile, the human visual masking theory hasbeen integrated into the metric of temporal flickering identification.Extensive experiments show that the proposed methods have a good performance.