| With the rapid development of information technology, the communication of video and video occupies the leading position of multimedia information. The main content of the mobile terminal video service such as high definition television, distance education and video conference has been widely spread and application. Since the 1970 s, the frame rate conversion technology has been used in the conversion of different frame rate video format, the temporal redundancy reduction in video coding and the improvement of flat-panel TV image quality. According to the research status and background of the frame rate up conversion, the existing deficiencies of the current frame rate up conversion algorithms has been compared and analyzed. Then, the frame rate up conversion algorithm which is based on the combination of forward & backward motion estimation and 3DRS(3-D Recursive Search) has been proposed. Compared with the traditional algorithm, this algorithm has more accurate motion vector, solves the overlap & hole problem and avoids the halo effect.Two kinds of frame rate up conversion algorithm has been studied. The non motion compensation frame rate up conversion algorithm has low computational complexity, easy realization and good performance in slow sequences, but it proves to be bad in fast sequences. However, the motion compensation frame rate conversion algorithm predicted intermediate frame in the neighboring frames using motion information. Thus, it is possible to show the real movement of objects, so the proposed frame rate conversion algorithm will use the latter. In addition, the proposed algorithm has been divided into three parts: motion estimation, motion vector post-processing and motion compensation.In the part of motion estimation, one-way motion estimation is accuracy but will affect the result of the interpolation. To order to solve the following issue, the forward & backward motion estimation has been proposed, and the obtained motion vector has been further processed in the motion vector post-processing part. Since the traditional block matching is easy to cause the block effect, the adaptive variable block method has been used in motion estimation so that the inaccurate block can be taken to further estimation. In order to improve the spatial & temporal correlation of the candidate motion vector, the selection of the candidate motion vector has been optimized. Since the candidate vector is likely to be fractional, the sub pixel motion estimation has been applied. At last, in order to improve the matching criterion of the traditional 3DRS algorithm, the concept of gradient has been introduced.In the part of motion vector post-processing and motion compensation, in order to solve overlaps and holes in the last part, the motion vector field has been retiming and the bi-directional motion vector can be obtained. Moreover, the complicated traditional median filtering algorithm has been improved due to its complicated calculation. In order to avoid halos, occlusion has been divided into three categories. At last, the adaptive interpolation can be done according to the classification.In the experiment, the subjective quality of the proposed algorithm can be obtained by observing the similarity between the original frame and the interpolation frame of this one and the other traditional algorithm. It can be obtained that the proposed algorithm is the best one. In the objective aspect, the curves of peak signal-to-noise ratio, mean structure similarity and sum of absolute motion neighboring difference which are generated by the proposed algorithm and other traditional algorithms has been compared respectively. Moreover, the mean values of peak signal-to-noise ratio and structure similarity according to the proposed algorithm and other traditional algorithms in all testing cases has been worked out. Finally, the comprehensive performance of the proposed algorithm is better than the other traditional algorithm such as frame average algorithm, 3DRS algorithm and full search algorithm. |
PDF Full Text Request