| Frame rate conversion (FRC) is a conversion process between any two displayformats with different frame rates. Due to the rapid development of advanced televisionand multimedia techniques, it has been widely used in many applications includingHDTV(High Definition Television) and low bandwidth video coding, and been recentlya new hot spot in the domain of video processing.As non-motion compensated frame rate conversion causes motion jerkiness andblurring at an interpolated frame, most of FRC use motion compensation algorithms. Insummary, motion compensated frame rate conversion consists of two major steps:motion estimation and motion compensation, and the work of this thesis focuses onmotion compensation in the realization of FRC. First, based on the existing motioncompensated interpolation algorithms, the thesis proposes a new interpolation algorithmtaking use of bidirectional motion vectors. Compared to traditional algorithms whichonly use single directional vectors, the proposed algorithm can get a better intermediateframe. General interpolation algorithms do not apply to nonlinear motion (e.g. complexmovements and occlusion, etc), for they are put forward in the situations where objectsmove linearly. Then based on occlusion analysis, the thesis adds two new additionalalgorithms to motion compensation in order to further improve the performance of theproposed interpolation algorithm. One of the two improvements is vectors refinement tomake vectors more accurate by correcting some false vectors. Another improvement isthe special interpolation algorithm aimed at occluded blocks in the interpolated frame:detecting whether there are two objects moving in different directions in the block of theintermediate frame, segmenting the detected occluded block basing on pixels, assigningcorresponding vectors to different segments, and selecting the proper reference framesto interpolate the segments. Experimental results show that the proposed algorithm canprovide good interpolated frame quality. |
PDF Full Text Request