Yanfei Liu, Fengchun Tian, Ran Liu. A Novel Approach to Frame Rate Up-conversion Based on Morphing. Journal of ComputationalInformation Systems, 2011, 7(14): 5219-5226 A Novel Approach to Frame Rate Up-conversion Based on Morphing All rights reserved. Do not used for commercial purposes.
5220 Y. Liu et al. /Journal of Computational Information Systems 7:14 (2011) 5219-5226
respect to the other frame[1, 2]. As unidirectional motion estimation will cause overlaps and holes, a number of researchers have proposed to use bi-directional motion estimation which divides the frame to be interpolated into blocks[3, 4]. Bi-directional motion estimation has become main approach of FRUC at present.
However, BMA is based on the assumption that all motions are translational for computational simplicity, so this approach can not model more complex motion such as rotation and scaling. Besides, the estimated motion vectors simply using BMA are often unreliable for frame interpolation in the sense that they fail to represent true motion. Directly employing these MVs usually results in unpleasant artifacts such as block artifacts, ghost artifacts in the interpolated frames. Some approaches can be adopted to remove block artifact and motion blur of BMA. For instance, Overlapped Block Motion Compensation(OBMC) technique is used to reduce block effects in[3], S. Sekiguchi employed vector median filter to remove motion outliers[5, 6], image and motion field segmentation techniques are used to refine the estimated motion vector field in[7], Yang Ling proposed a MV refinement method according to both the spatial correlation in the interpolated frame and the temporal correlation in the previous and current frame[8]. However, there are also inevitable defects in BMA, for example, estimated motion vector fields are not continuous and it cannot model complex motion such as rotation and scaling. Therefore, there is an insistent demand on a new method to solve these problems thoroughly. In this paper, we address a novel FRUC algorithm based on morphing technique to overcome the problem of BMA. The proposed method employs morphing techniques to generate smooth transitions between images.
The remainder of this paper is organized as follows. In section 2, the proposed FRUC method for are presented. Section 3 extends the method to implement prewarping. The experimental results and comparison with conventional methods are shown in section 4. Finally, a conclusion is provided in section 5.
2. Proposed Algorithm
The overall framework of proposed FRUC algorithm is depicted in Fig. 1. At the beginning, we read two reference frames: previous frame and current frame, from video sequence, and calculate the variance of the current frame to estimate if this video sequence has ample texture. If the variance is greater than a predefined threshold, the current video is identified as an ample-textured one and a motion estimation algorithm without prewarping is used to match feature points. This is because prewarping is less effective for ample-texture video but motion estimation is more effective on this kind of video. If variance is not greater than the threshold, an algorithm with prewarping step will be adopted. The two different methods to generate corresponding feature points of the two frames are depicted in the following parts 2.1 and 2.2 respectively. Finally, depending on the corresponding points we use Delaunay Triangulation(DT) to generate triangle mesh and use linear interpolation to produce new frame, which can be interpolated into original video sequence to create frequency-doubled video sequence. DT method connects adjacent feature points by measuring Euclidean distance, which is consistent with the demand of image correlation. Meanwhile DT mesh avoids long-narrow and sharp meshes as much as possible. Therefore, we adopt DT algorithm to create triangle mesh. As for the vertices of triangle mesh, we adopt linear interpolation. And for the points within some triangle, we use the relative position coordinate of the point to linearly interpolate its pixel value.
compensated frame rate up-conversion (MC-FRUC) is the best approach [5]....a posteriori probability (MAP) estimation based on the Bayesian framework [...
The main contribution of our work is the novel ...morphing method to create a normal frame-rate ...If it is, we perform a feature-based image ...
Mbps at a frame rate of 8.5 frames per ...from new viewpoints by morphing live video streams...A hardware accelerated approach to high quality point...
Mbps at a frame rate of 8.5 frames per ...from new viewpoints by morphing live video streams...A hardware accelerated approach to high quality point...
热门文档