Research On Key Technology In Video Image Pre-processing

Video pre-processing technology is an image processing technology which employees avariety of pre-processing algorithm to obtain a better video image quality and high compressionratio of video. With the rapid development of multimedia display technology, it plays anincreasingly important role in video image processing and application. This paper has furtherstudied the de-interlacing and noise reduction in video pre-processing. To save bandwidth andreduce the large area of the image shake, interlaced scanning is a widely used scanning format.Progressive-interlaced display system will bring the flash, screen shaking, teeth and other problems.With the widespread use of large-screen TV in a variety of occasions, the problems whichinterlaced scanning takes are becoming increasingly prominent. So, it is necessary to study effectivede-interlacing algorithms. In addition, since the actual scene image sensor converts the light signalto image signal, there is always a variety of noise. The noise not only increases the entropy of theoriginal sequence, but also reduces the efficiency of the encoder. Because noise is uncorrelated intime domain, the encoder needs more bits to encode, which causes the degradation of efficiency.Therefore, the reduction of video images is necessary.In this paper, we have in-depth studied the de-interlacing and noise reduction technology invideo image pre-processing, and proposed improved de-interlacing and noise reduction algorithmsrespectively. This paper is organized as follows:Firstly, this paper describes the video image quality evaluation criteria and analysis theimportance of the video image quality evaluation criteria applied in video image. Then we use thestructural similarity index metric (SSIM), average resolution criteria and the peak signal to noiseratio (PSNR) to evaluate the algorithm for de-interlacing, and PSNR for de-noising algorithm.Second, we analyze the de-interlacing techniques of the video image pre-processing, includingclassical linear and nonlinear algorithms, and discuss the advantages and disadvantages of variousalgorithms. On this basis, this paper presents a video de-interlacing algorithm based on the medianfilter and edge interpolation and the algorithm has been simulated. Experimental results show thatthe algorithm effectively protects horizontal, vertical and diagonal edges of the image, andeffectively solve the afterimage phenomenon, meanwhile simulation data are improved. We utilizeSSIM, average resolution criteria and PSNR to evaluate de-interlacing algorithm and the proposedalgorithm is superior to the line average, edge-based line average (ELA), NEW ELA andthree-point median algorithm with motion detection.Thirdly, we introduce the space domain and transform domain de-noising method of the videoimage briefly. Traditional method of de-noising is accomplished before encoding the video image to remove noise, in this paper, we integer the local statistics filter in the space domain and H.264coding framework to present a new video de-noising algorithm in time domain and we haveachieved the algorithm. Simulation results show that: the time domain filtering has a better result inboth filtering effects and the compression performance than the space domain, at the same time, theproposed algorithm has a better protection for the details of the image.