| Abstrakt: |
In daily life, video blurring is often caused by camera shake or damage of physical components, which in turn affects the target segmentation effect in the video. Therefore, in order to solve the problem of video deblurring and target segmentation. In this research, based on the scale recurrent network, Hal 2D wavelet transform and attention mechanism are introduced, and the obtained algorithmic model is tested for performance in a dataset, and at the same time compared with the previous traditional algorithmic model to screen out a more effective deblurring algorithm. Based on the traditional video target segmentation algorithm, the attention mechanism and morphological module are introduced and a new optimisation algorithm for video target segmentation is proposed. The performance of this algorithm is compared with the previous traditional algorithm model, and a more effective target segmentation algorithm is screened. Finally, the screened two optimal algorithms are combined to compare with the previous traditional video deblurring and target segmentation algorithms using basketball videos as experimental objects. The experimental results show that the peak-to-noise ratio of this new deblurring algorithm is as high as 30.55, and the structural similarity is as high as 0.942. The new segmentation algorithm has a region similarity as high as 96.5%, and the contour accuracy is as high as 93%. The processing time of the algorithm is shorter and the effect is better when the two algorithms are combined. In summary, the novel algorithm provides a certain reference value for the further research of subsequent video processing technology. [ABSTRACT FROM AUTHOR] |
