Film-to-Video Conversion and Hybrid Interlaced-to-Progressive Conversion Algorithms with Reliable Interfield Information

碩士 === 東海大學 === 資訊工程與科學系 === 94 === In this paper, we discussed about two issues. One is film-to-video frame rate conversion, the other is deinterlacing technology. In film-to-video frame rate conversion, the scene-cut detection is executed first in every four frames. Using histogram-based and p...

Full description

Bibliographic Details
Main Authors: Ming Che, Chen, 陳銘哲
Other Authors: Ching Trsong, Tsai
Format: Others
Language:zh-TW
Published: 2006
Online Access:http://ndltd.ncl.edu.tw/handle/27480621935539578611
Description
Summary:碩士 === 東海大學 === 資訊工程與科學系 === 94 === In this paper, we discussed about two issues. One is film-to-video frame rate conversion, the other is deinterlacing technology. In film-to-video frame rate conversion, the scene-cut detection is executed first in every four frames. Using histogram-based and pixel-based statistic methods to judge whether two consecutive frames have been scene-cut or not. According the detection result, the motion compensation interpolation will generate a new frame in different display position. Our method shows a lower computation and better PSNR value than other methods. In second issue, we proposed a hybrid deinterlacing system. We induce five scene-changed phenomenons. By computing Motion Pixel Amount(MPA) and Motion Block Amount(MBA) in the same parity frames, we can consider whether scene changed or not. If scene changed, we will use motion adaptive deinterlacing, 2D-ELA, to interpolate generated pixels in the slots of current frame. Otherwise, the motion compensation deinterlacing will be executed. This method can successfully avoid the artifacts distortion while scene changed. And it also shows better PSNR value than other methods in average.