ing at the applications of moiré fringes in fine measurement, a super-resolution algorithm is proposed based on local self-similarity and deblocking post-processing. In this algorithm, with the local self-similarity of moiré fringes, the initial high-resolution images are first obtained through an interpolation of the original low-resolution images. Then the optimal matching low-resolution block corresponding to each high-resolution image block is searched. The prior knowledge is extracted from the high- and low-resolution image blocks and thus the super-resolution reconstruction of a single-frame image is realized. In addition, the blocking artifacts are introduced in the reconstructed results after the blocking operations. As for this problem, a post-processing algorithm for quickly eliminating blocking artifacts is simultaneously proposed. The results show that the combination of the proposed two algorithms can effectively improve the image quality and simultaneously eliminate the blocking artifacts in the reconstruction process of images. The algorithm does not relay on external images and has a low computational complexity, suitable for the super-resolution reconstruction of moiré fringe images.

ing at the problems of ghost and the noise interference from dynamic background in classic visual background extraction algorithm, an improved visual background extraction algorithm is proposed. The important feature information of pixels in complex environment can be collected by creating the auxiliary sample set. By introducing analysis of the pixel ghost factor and the region complexity, the matching threshold and updating rate of each pixel can be adaptively adjusted. With the pixel flicker analysis based on sliding window, the points which may be misdetected as foreground can be updated to the auxiliary samples according to probability. The comparative experiments in the multi-scene show that the proposed method can reduce the wrong classifications rate to as low as 1.49%, eliminate the ghost quickly, suppress the noise interference from the dynamic background, and ensure the complete recognition of foreground target. The results of the algorithm are more accurate in the complex environment.

ing at the problem of tracking failure due to fast motion and scale variation during object tracking, an object scale adaptation tracking based on full-convolutional siamese networks is proposed. First, a full-convolutional symmetric network is constructed using MatConvNet framework, and the multidimensional feature maps of template images and experimental images are obtained by using the trained networks. Through the cross-correlation operation, the point with the highest confidence score is selected as the center of the tracked target. Then, through multi-scale sampling at the center, the error samples that are less than half the template variance are filtered out. The probability histograms of target templates and samples are built. The Hellinger distance between the template and the samples is calculated, and the appropriate scale is selected as the scale of the target tracking window. Experiments on the OTB-13 dataset are carried out. Compared with other tracking algorithms, the tracking success rate of proposed method is 0.832, and the precision is 0.899, which are higher than that of other algorithms, and the average tracking speed is achieved 42.3 frame/s, meeting the needs of real-time object tracking. Selecting the tracking sequences with fast motion or scale change attributes for further testing, the tracking performance of proposed method is still higher than other algorithms.