Ke Xiang

Robust visual tracking via CAMShift and structural local sparse appearance model

A new background analysis method is proposed based on sparse representation.We propose a semi-global weighted search algorithm for the local optimization problem.We develop a model update strategy for long-term robust tracking. This paper addresses issues in visual tracking where videos contain object intersections, pose changes, occlusions, illumination changes, motion blur, and similar color distributed background. We apply the structural local sparse representation method to analyze the background region around the target. After that, we reduce the probability of prominent features in the background and add new information to the target model. In addition, a weighted search method is proposed to search the best candidate target region. To a certain extent, the weighted search method solves the local optimization problem. The proposed scheme, designed to track single human through complex scenarios from videos, has been tested on some video sequences. Several existing tracking methods are applied to the same videos and the corresponding results are compared. Experimental results show that the proposed tracking scheme demonstrates a very promising performance in terms of robustness to occlusions, appearance changes, and similar color distributed background.

Attention shift-based multiple saliency object segmentation

Abstract. Object segmentation is an important but highly challenging problem in computer vision and image processing. An attention shift-based multiple saliency object segmentation model, called ASMSO, is introduced. The proposed ASMSO could produce a pool of potential object regions for each saliency object and be applicable to multiple saliency object segmentation. The potential object regions are produced by combing the methods of gPb-owt-ucm and min-cut graph, whereas the saliency objects are detected by a visual attention model with an attention shift mechanism. In order to deal with various scenes, the model attention shift-based multiple saliency object segmentation (ASMSO) contains different features which include not only traditional features, such as color, uniform, and texture, but also a new position feature originating from proximity of Gestalt theory. Experiments on the training set of PASCAL VOC2012 segmentation dataset not only show that traditional color feature and the proposed position feature work much better than features of texture and uniformity, but also prove that ASMSO is suitable for multiple object segmentation. In addition, experiments on a traditional saliency dataset show that ASMSO could also be applied to traditional saliency object segmentation and performs much better than the state-of-the-art method.

Efficient local and global contour detection based on superpixels

Abstract In this paper, two contour detection methods, inspired from gPb framework, are introduced and applied to saliency object segmentation. To improve the computational efficiency of gPb method, superpixels are introduced into the computational processes of both mPb and sPb. Specifically, for mPb, only the pixels within a given distance from the boundaries of superpixels are considered. For sPb, graph is constructed from superpixels and some selected pixels. Experiments on a public available BSDS500 image dataset show that higher efficiency could be achieved by the proposed local contour detection method, mPbSP, than mPb while with competitive results. Besides, compared with state-of-the-art methods, better results could be produced by the proposed global contour detection method, gPbSP, when a relatively small distance is considered. Moreover, experiments on PASCAL VOC2012 training segmentation dataset show that competitive results of saliency object segmentation could also be produced by the proposed methods with much less time.

Optimal bi-directional seam carving for compressibility-aware image retargeting

Abstract A novel compressibility-aware image retargeting method based on seam carving is introduced in this paper. We propose a new significance detection method, with both the edge information and visual saliency taken into consideration. A wall-seam model is constructed to evaluate the image compressibility and assign the right number of seams for each direction. By repeatedly carving out or inserting seams we can retarget the image to a new size while preserving the important content. Finally, our algorithm is completed with the supplement of uniformly scaling, the stretched image is resized to the target size with the least structure distortion brought. Experimental results on images show that those improvements are effective and our approach can preserve image content better compared to several state-of-the-art image retargeting methods.

Video stabilisation based on modelling of motion imaging

This study proposes a novel digital video stabilisation scheme based on modelling of motion imaging (MI). The modelling of MI eliminates the speed motion as a result of a moving car, which is ignored in other models such as rotation + translation model, and estimates movement parameters of the background in video sequences captured from cameras mounted on moving cars. The authors first analyse the MI to understand the principle of the effects of car motion on MI, and select the matching method according to the proposed model. Then, they employ symmetric points to remove the speed motion. Finally, unwanted motion vector is stimulated by employing adaptive step-length filter, and the boundary compensating approach is employed to suppress the image jitter effectively. Their major contribution is the elimination of the effect of carriers speed in motion estimation. Other contributions include new robust block matching approach and adaptive-step selection for motion filtering. They conduct experiments on real videos and artificial data. Experiments on real videos show that the proposed model can remove the effect of car motion, whereas the experiments on artificial data are conducted for theoretical analysis.

Random walks colour histogram modification for human tracking

Accurate human tracking in surveillance scenes is one of the preliminary requirements for other tasks. However, when the human target is small, the extracted features may not be prominent and thus the tracking performance is unsatisfactory. The colour feature is relatively robust to the change of target size and shape, but it is prone to be affected by the background information. For the above reasons, the authors introduce random walker segmentation into human tracking and determine the background region according to the distribution characteristics of segmentation results. Even if the colour of the target is very similar to that of the background, this algorithm can segment the target. Furthermore, the principal component analysis method is used to distinguish human targets from the background as well. During tracking, the authors prevent the degradation of the target model by adding new target information. In order to overcome the mean-shift local optimisation problem, the authors search for the candidate target region with the largest weight according to the sum of all probability in each region. Experimental results further show that the authors’ tracking algorithm demonstrates better performance on tracking small human target under some challenging scenes compared with several existing tracking methods.

Face Recognition based on scale invariant feature transform and Spatial Pyramid Representation.

For Face Recognition (FR) task, local feature based approaches have proven to be more robust to variations than the holistic approaches. As one of the state-of-the-art local descriptors with excellent performance for matching different images of an object or a scene, Scale Invariant Feature Transform (SIFT) has sparingly been used in FR, and whether it is a good descriptor for face images needs to be analyzed more. In this paper we propose a face representation approach based on SIFT and Spatial Pyramid Representation (SPR), to enhance the recognition performance of SIFT based method. Firstly it employs SIFT to extract discriminative local features and then constructs Spatial Pyramid to form a local-holistic representation of a face image. Finally Classifiers such as Nearest Neighbor (NN) and Support Vector Machine (SVM) could be performed on representation vectors of equal length. The comparative experimental results on ORL and Yale databases indicate that our approach achieves better performance than other SIFT based methods. In addition, it shows great robustness against environmental variations such as pose mismatches, imperfect face alignment, various facial expressions and accessory configuration variations, demonstrating a new alternative method for FR task.

An experimental comparison of superpixels detection methods for contour detection

Recently, many superpixels detection methods have been proposed and used in various applications. We are interested in which method is more suitable for the application of contour detection. In this paper, superpixels are evaluated on BSDS500 dataset in two different aspects. On the one hand, contours are directly provided by the boundaries of superpixels and experiments show that better results could be achieved by the superpixels with irregular shapes than those with regular shapes and similar sizes. On the other hand, contours are further detected from those candidate positions which are confirmed by the boundaries of superpixels through the operation of dilation. In this situation, experiments show that competitive results could also be achieved by some superpixels with regular shapes and similar sizes. Besides, we propose a superpixels detection method called watershed-based graph (WG), by which superpixels with irregular shapes could be produced. Firstly, a graph is constructed from an over-segmented map which is achieved through a watershed algorithm. Then, to get the desired superpixels, the graph is segmented by merging neighbor segments in an order of decreasing similarity. Experiments show that higher efficiency could be achieved by WG with a moderate worse contour quality than its original graph-based method.

Spatiotemporal image edge analysis method for motion detection under the pan-tilt camera

Abstract. According to the background, we divide motion detection into two categories: the detection under static background (static camera) and the detection under moving background (moving camera). The motion detection algorithm under a moving camera is more complex than that under a static camera due to the mixture of the motions of target and background. We propose a spatiotemporal image edge analysis method (STIEA) for motion detection under a pan-tilt (PT) camera. First, we obtain the spatiotemporal image (STI) by extracting the corresponding rows from each frame and displaying them one above another after motion estimation. Subsequently, the distribution characteristic (the slope of each edge) of STI is further analyzed theoretically. Overall, the pixels with the same gray value in STI are almost distributed in the same line (linear characteristic). Based on this characteristic, we cluster the edges and detect the abnormal edges. The moving objects exist in the columns with abnormal edges. In addition, the influence of parallax on edges slope is also analyzed. Finally, our method is compared with several existing algorithms. The experimental results demonstrate that our algorithm performs better than others in various scenarios, which provides a unique thought for motion detection under a PT camera.

Ensemble Simplified Fuzzy ARTMAP with Modified Plurality Voting

In this paper, a method estimating the probability which is used in plurality voting is employed to work for plurality voting. The unknown probability in plurality voting leads to poor performance when classification is performed. Confusion matrix could estimate the probability effectively which is employed in this paper to estimate the probability. UCI datasets have been used to evaluate the applicability and performance of the proposed method. The experiments demonstrate the potential of the proposed method if offering an optimal solution to the plurality voting.