Reseach And Implementation Of The Deinterlacing And Denoising Algorithm On HD Video

With the development of multimedia technologies,the digital TV systems and display equipment on the video signal quality requirements are getting higher and higher,video post-processing technology has become increasingly important.The deinterlacing and denoising algorithm is becoming indispensable for the video post-processign technology.The deinterlacing technique is intended to solve the problem of jagged effects due to interlaced scanning.While the denoising algorithm is improving the quality of the video signal,it also becomes preprocessing process for the other video-enhancement algorithm.Therefore,the study of the denoising has strong practical significance.The algorithm of high computational complexity can encounter the bottleneck of over time in reality and can not be used widely.So the parallel design of video processing algorithms has a high practical significance.In this thesis,an edge adaptive deinterlacing algorithm based on motion detection is proposed for deinterlacing.In order to solve the problem of high complexity of BM3D denoising algorithm,a parallel program is designed.The main work of this thesis is as follows:(1)Because of the jagged effects of the interlaced video,this thesis proposes an improved deintercing algorithm based the motion detection.This thesis introduces 5-fields information to detect the motion information more precisely for the sake of more related fields' information.This thesis introduces a method of dynamic threshold and the value of the threshold is changing with the gray value of the pixel.This reduces the detection error effectively.(2)After getting the motion information,an interpolation method to the field information can help us get the frame information.The thesis introduces a texture detection before beginning the interpolating.The texture detection can distinguish the smooth and texture regions.For the two different parts,we use different interpolation method and we can get more texture information and high frequency component.Experiments show that the algorithm in this thesis can get a better deinterlacing effects.(3)The BM3D denoising algorithm is one of the best one for denoising.But because of the high computational complexity,it only produces one frame during sereral seconds and this limits the application in reality.Because of that the thesis designs a parallel program and from the experiment comparison it get a high acceleration ratio.