Zhaoquan Cai

Improving sampling-based image matting with cooperative coevolution differential evolution algorithm

Image matting is a fundamental operator in image editing and has significant influence on video production. This paper explores sampling-based image matting technology, with the aim to improve the accuracy of matting result. The result of sampling-based image matting technology is determined by the selected samples. Every undetermined pixel needs both a foreground and background pixel to estimate whether the undetermined one is in the foreground region of the image. These foreground pixels and background pixels are sampled from known regions, which form sample pairs. High-quality sample pairs can improve the accuracy of matting results. Therefore, how to search for the best sample pairs for all undetermined pixels is a key optimization problem of sampling-based image matting technology, termed “sample optimization problem.” In this paper, in order to improve the efficiency of searching for high-quality sample pairs, we propose a cooperative coevolution differential evolution (DE) algorithm in solution to this optimization problem. Strong-correlate pixels are divided into a group to cooperatively search for the best sample pairs. In order to avoid premature convergence of DE algorithm, a scattered strategy is used to keep the diversity of population. Besides, a simple but effective evaluation function is proposed to distinguish the quality of various candidate solutions. The existing optimization method, original DE algorithm and a popular evolution algorithm are used for comparison. The experimental results demonstrate that the proposed cooperative coevolution DE algorithm can search for higher-quality sample pairs and improve the accuracy of sampling-based image matting.

Multiple human tracking based on distributed collaborative cameras

Due to the horizon limitation of single camera, it is difficult for single camera based multi-object tracking system to track multiple objects accurately. In addition, the possible object occlusion and ambiguous appearances often degrade the performance of single camera based tracking system. In this paper, we propose a new method of multi-object tracking by using multi-camera network. This method can handle many problems in the existing tracking systems, such as partial and total occlusion, ambiguity among objects, time consuming and etc. Experimental results of the prototype of our system on three pedestrian tracking benchmarks demonstrate the effectiveness and practical utility of the proposed method.

A Novel Semi-fragile Watermarking Algorithm for 3D Mesh Models

A novel semi-fragile watermarking algorithm based on mesh content was proposed. 3D mesh models were calibrated by principal component analysis first, and then this algorithm constructed spherical coordinates mapping square-matrix to realize 2D parameterization of vertexs geometric data. A binary image was chosen as the semi-fragile watermark and hidden in coefficients of square-matrix. Experimental results show that the proposed algorithm can tolerate various mesh normal processing, such as mesh translation, mesh rotation, mesh uniformed scaling, while is sensitive to malicious tampering, such as vertex random noising.

Improving Sample Optimization with Convergence Speed Controller for Sampling-Based Image Matting

Image matting is a core and challenging operator when processing images or videos. Its aim is to accurately extract the foreground region from an image. In this paper, we explore sampling-based image matting. The key optimization problem of sampling-based image matting is how to search the best foreground-background sample pair for every undetermined pixel. It is termed as the “sample optimization problem”. Many sample optimization algorithms have been proposed for improving the efficiency of searching the best foreground-background sample pair. However, they fail when premature convergence is occurred. This paper presents a new sample optimization algorithm, which is based on convergence speed controller (CSC). The CSC is a general algorithm strategy. It can be embedded into algorithms and enhance the performance of the algorithms by maintaining the convergence speed and preventing premature convergence. By comparing with existing sample optimization algorithms, the experimental results show that our algorithm is competitive and effective to search the best sample pair and improve the performance of sampling-based image matting.

Offline Video Object Retrieval Method Based on Color Features

At present, video retrieval has been applied to many fields, for example, security monitoring. With the development of the technique of content-based video retrieval, video retrieval will be applied to more areas. The article mainly do research on offline video retrieval based on color features and realize offline video color features retrieval. The research realized Algorithm for Video Objective Tracking based on Adaptive Hybrid Difference and was focused on designing color features range calculation scheme with the combination of RGB and HSL color model. And extract and judge the color feature of the blob in the video then analyze and process the retrieval result. According to the result of this test, the success rate of detection of the system have reached ninety percentage upon. The realization of offline video object retrieval system based on the color features can decrease the time of Manual Retrieval to the color features object in the video, help people filter information and have benefits on the realization of intelligent and automatic video retrieval.

Particle Swarm Optimization with Convergence Speed Controller for Sampling-Based Image Matting.

Image matting is a challenging task and has become the basis of various digital multimedia technologies. The aim of image matting is to extract the foreground from a given image with the user-provided information. This study focuses on sampling-based image matting methods. The key issue in sampling-based image matting methods is to search the best foreground-background (F-B) sample pair for each unknown pixel which is generally known as a large-scale “sample optimization problem’’. This study explores a new variant particle swarm optimization algorithm based on convergence speed controller, a premature-convergence-prevented strategy, to improve the performance of image matting. Particularly, we embed the convergence speed controller into particle swarm optimization and proposed a efficient variant algorithm of it for the sample optimization problem. We conducted extensive experiments to verify the efficiency of the proposed algorithm. The experimental results show that the proposed algorithm, compared to the existing algorithms, is competitive and can achieve higher-quality matting.

A discrete bio-inspired metaheuristic algorithm for efficient and accurate image matting

With the development of digital multimedia technologies, image matting has become one of the most popular research problem in academic field and been widely applied in industrial communities. The key challenge of image matting is how to extract the foreground region (target region) from a given image accurately. Sampling-based image matting technology implements matting by sampling some foreground pixels and background pixels from known regions and finding the best foreground–background sample pair for every undetermined pixel. The best foreground–background sample pair represents the true foreground and background colors of the corresponding undetermined pixel and they can estimate the region of this undetermined pixel accurately. Therefore, the quality of matting depends on whether the best sample pair can be found. This search process can be regarded as a combinational optimization problem. In this paper, in order to obtain more accurate matting result, we applied a bio-inspired metaheuristic algorithm to solve this problem, which is based on the promising earthworm optimization algorithm (EWA). By analyzing the property of this optimization problem, we upgrade two reproductions and the cauchy mutation of EWA to discrete calculations. The proposed algorithm is called as the discrete earthworm optimization algorithm (D-EWA). By comparing with existing optimization algorithms on a standard benchmark dataset, the experimental results show that the proposed D-EWA can obtain more accurate matting results on both visual effect and quantitative metric.

Unsupervised segmentation evaluation: an edge-based method

Unsupervised segmentation evaluation method quantifies the quality of segmentation without the reference segmentation or user assistance. Although some methods have been proposed to statistically analyze the pixel values, these methods are not sensitive enough to provide a metric of segmentation quality. This paper uses the image edge, a more robust feature, to measure the quality of segmentation. An edge-based segmentation evaluation method is introduced in this paper, which can be applied to both image and single region segmentation evaluation. The proposed method evaluates the quality of segmentation with three edge-based measures: the edge fitness, the intra-region edge error, and the out-of-bound error. These measures encourage the outline of segmentation to align with the edge and punish the segmentation that exceeds the edge. Experiments results show that our method is more sensitive to under-segmentation and over-segmentation. Using the parameters optimized by the proposed method, the segmentation produced by the classic region growing method is visually similar to the state-of-the-art segmentation method.

An hybrid evolutionary algorithm with scout bee global search strategy for Chinese Nurse Rostering Problems

Nurse Rostering Problem (NRP) is one of NP - hard combinatorial optimization problems about the distribution of medical resources. In the past, there have been several proposed methods like heuristic algorithms and algorithms based on establishing rigorous mathematical models. Especially, the hybrid algorithm combined integer programming and evolutionary algorithm (IP+EA) have been proved to be effective for NRP. However, these methods are not efficient in dealing with large-scale NPR instances, like Chinese NRP. In order to overcome the premature convergence of IP+EA, we propose a hybrid evolutionary algorithm based on scout bee global search strategy. Inspired by the behavior of scouts in artificial bee colony algorithms, the global search is integrated into EA, which can lead the algorithm to escape from local optima. The experimental results indicate that, our proposed approach is more effective than several existing algorithms to solve the Chinese NRP.

Edge-Based Unsupervised Evaluation of Image Segmentation

This paper focuses on a fundamental problem in computer vision: how to evaluate the quality of image segmentation. Supervised evaluation methods provide a more accurate evaluation than the unsupervised methods, but these methods cannot work without manually-segmented reference segmentations. This shortcoming limits its applications. We present an edge-based evaluation method which works without the comparison with reference segmentations. Our method evaluates the quality of segmentation by three edge-based measures: the edge fitness, the intra-region edge error and the out-of-bound error. Experimental results show that our method provides a more accurate evaluation than those method based on the statistic of pixel values, and can be used in both segmentation evaluation and region evaluation. A significant linear correlation is shown between the evaluation scores of our method and two widely used supervised methods. The proposed methods show a high performance on the automatic choice of the best fitted parameters for region growing.