Search

In the realistic video monitoring environment, the traditional super-resolution reconstruction technique based on prior knowledge is not suitable for monitoring the super-resolution reconstruction of the image. In this paper, a super-resolution reconstruction algorithm of target image based on learning background is proposed. The first part of the algorithm is to design a non-manifolds consistency algorithm for super-resolution reconstruction of the whole video surveillance image. The second part of the algorithm, from video surveillance images in the background, to select the characteristics significantly, and the relatively fixed background. And then to study the background, study a mapping function can improve image quality. Finally, the mapping function to restoration image of interested target, so that we can better recover the structure and texture of target image details. The experimental results show that the proposed algorithm improves both the objective evaluation index and the subjective visual effect.

In the most of super-resolution reconstruction algorithms, high-resolution and low-resolution images are assumed in the same manifold space. However, due to distractions, this assumption is not suitable for the practical applications. This paper proposes a novel super-resolution reconstruction algorithm for face images. In order to consider the manifold inconsistency of high-resolution and low-resolution images, a mapping function revealing the relationship between the coefficients of high-resolution and low-resolution images is introduced during the stage of dictionary training. During the stage of image reconstructing, the sparse coefficients of high-resolution images can be calculated by mapping function with the input of sparse coefficients of low-resolution images. To remove the irregularity of image edge of initial high-resolution images, an optimization process is presented based on the self-similar characteristic of face images. Experimental results show that the proposed algorithm can perform better in PSNR and SSIM.

It is a challenging task to develop a robust appearance model due to various factors such as partial occlusion, fast motion, background clutters and illumination variations. In this paper, we propose a novel target representation for visual tracking. Namely, a target candidate is represented by sparse affine combinations of dictionary templates in a particle filter framework. Affine combinations based target appearances can cover unknown appearances. In order to adapt the dynamic scenes across a video sequence, the dictionary templates are updated in the tracking process. Experimental results on several challenging video sequences against some state-of-the-art tracking algorithms demonstrate that the proposed algorithm is robust to illumination variations, background clutters, etc.

Developing an effective target appearance model is a challenging task due to the influence of factors such as partial occlusion, illumination variations, fast motion, etc. Existing appearance models usually utilize the tracking results from previous frames as target templates upon which the target appearance model is built by linear combinations of the templates. With such kind of representation, visual tracking is not robust when drastic appearance variations occur. We propose a simple but effective tracking algorithm with a novel appearance model in a particle filter framework. A target candidate is represented by the convex combination of a set of target templates. Additionally, the distance between a target candidate and the templates is measured using the EMD. Experimental results on challenging video sequences against state-of-the-art algorithms demonstrate the robustness and effectiveness of the proposed tracking algorithm.

Gaussian Mixture Model (GMM) and its variations process images by per pixel, so they may be corrupted by noises and the computational cost is high. In this paper, we propose a robust moving object detection algorithm with a background dictionary learning. To do this, we first divide an image into multiple image patches that have the same sizes. Each patch is the object or background. Then, A background dictionary is learnt for each patch. The similarity between a patch and the background dictionary is measured, upon which a patch is distinguished between the object and the background. Additionally, in order to adapt the dynamic contexts across in a video sequence, a robust background dictionary updating scheme is proposed. Experimental results demonstrate the effectiveness and robustness of the proposed detection algorithm.

This paper proposes a vehicle detection and traffic volume statistics algorithm Based on an improved single Gaussian model. This algorithm contains three major sections, which are moving target detection, shadows suppression and traffic volume count. Firstly, with an improved background initialization method, the paper detects the moving target by using the single Gaussian model. Then, a computational method of Shadows suppression is presented in the RGB feature space. Finally, the traffic volume is counted in the virtual areas of lanes. Two experiments are performed on the algorithm. The results show that the algorithm can detect vehicle quickly and have a higher recognition rate. It can works well in the actual conditions.

The key of intelligent infusion system is to detect liquid level of infusion bottle fast and accurately. We proposed an automatic detection algorithm of liquid level of infusion bottle based on image processing, by designing edge detection, binarization, filtering, image projection and motion detection to realize automatic liquid level detection. The algorithm has the advantages of stable noise-resistence, rapidity and simplicity, etc. Experimental results of the algorithm show that the algorithm is effective and feasible.

This paper address issues that arise in copyright protection systems of digital images, which employ blind watermark verification structures in the curvelet domain. First, we observe that statistical distribution with heavy algebraic tails, such as the alpha-stable family, are in many cases more accurate modeling tools for the curvelet coefficients than families with exponential tails such as generalized Gaussian. Motivated by our modeling results, we then design a new processor for blind watermark detection using the Cauchy member of the alpha-stable family. We analyze the performance of the new detector in terms of the associated probabilities of detection and false alarm and we compare it to the performance of the generalized Gaussian detector and the traditional correlation-based detector by performance experiments. The experiments prove that Cauchy detector is superior to the others.

This paper discusses a type of intelligent transfusion system based on computer vision, which key is the algorithm of image segmentation and liquid level detection. The template matching is adopted for image segmentation, and a fast and efficient automatic detection algorithm is proposed to detect liquid level by designing temporal difference, accumulative difference and projection analysis to realize automatic detect liquid level. The algorithm has the advantages of stability noise-resistant ability, rapid and simple, etc. Experimental results of the algorithm show that the algorithm is effective and feasible.

This paper discusses a type of intelligent transfusion monitor system based on image processing, which key is the algorithm of image segmentation and liquid level detection. The template matching is adopted for image segmentation, and a fast and efficient automatic detection algorithm is proposed to detect liquid level by designing orientation gradient and projection analysis. The algorithm has the advantages of stability noise-resistant ability, rapid and simple, etc. Experimental results of the algorithm show that the algorithm is effective and feasible.

A Chinese chessman pattern recognition system is presented. First, system structure is introduced, and chessman image pretreatment is discussed. Then primarily discusses the chessman recognition algorithms. Because of direction haphazardry, chessman recognition is quite different from character recognition. This paper presents two kinds of chessman recognition algorithms - RC and concentric circle algorithm for feature extraction. Then choose a relatively suitable recognition algorithm by comparing their time complexity and accuracy.

It is difficult to filter an image while protecting its detail. In this paper, a novel filter is proposed for detail preserving restoration of images. Corrupted pixels are detected by a multi-scale way. For the corrupted pixels, we use a kind of four direction interleaved interpolation value to replace their value. The performance of the proposed operator is tested for different test images corrupted at various noise densities. Experimental results show that the proposed filter can not only provide better performance of suppressing noise but also preserve more detail features. It is also demonstrated preferable performance in suppressing different impulse noise and random noise.