Zhiheng Zhou

Gradient descent with adaptive momentum for active contour models

In active contour models (snakes), various vector force fields replacing the gradient of the original external energy in the equations of motion are a popular way to extract the object boundary. Gradient descent method is usually used to obtain the equations of motion by minimising the energy functional. However, it always suffers from local minimum in extracting complex geometries because of non-convex functional. Gradient descent method with adaptive momentum term is proposed in this study. First, an acceleration function of evolution is defined. Then, the adaptive momentum term is obtained by calculating the product between the edge stopping function and the defined acceleration function. Finally, adaptive momentum is compatible with the snakes. The edge stopping function is used to decide the influence region of the momentum, whereas the defined acceleration function determines the magnitude of the momentum. It is used to extract the complex geometries (such as deep concavity) when adding the adaptive momentum into some snakes, such as gradient vector field or vector field convolution snakes. On the other hand, the proposed method also accelerates the rate of convergence. It can be applied to extract a single object in real images. The experimental results show that the proposed method is effective and efficient.

Global minimization of adaptive local image fitting energy for image segmentation

The active contour model based on local image fitting (LIF) energy is an effective method to deal with intensity inhomo-geneities, but it always conflicts with the local minimum problem because LIF has a nonconvex energy function form. At the same time, the parameters of LIF are hard to be chosen for better performance. A global minimization of the adaptive LIF energy model is proposed. The regularized length term which constrains the zero level set is introduced to improve the accuracy of the boundaries, and a global minimization of the active contour model is presented. In addition, based on the statistical information of the intensity histogram, the standard deviation σ with respect to the truncated Gaussian window is automatically computed according to images. Consequently, the proposed method improves the performance and adaptivity to deal with the intensity inhomo-geneities. Experimental results for synthetic and real images show desirable performance and efficiency of the proposed method.

Fuzzy Blocking Artifacts Reduction Algorithm Based on Human Visual System

Traditional blocking artifacts reduction algorithm often causes image edges losing. To address this problem, a fuzzy edge-sensitivity blocking artifacts reduction algorithm is proposed. Based on some characteristics of human visual system, this algorithm applies adaptive filter to current pixel by integrating fuzzy logic technique. The experimental results show that this algorithm can obtain better results than traditional ones.

Efficient adaptive MRF-MAP error concealment of video sequences

A new adaptive error concealment method based on Markov random field (MRF)-maximum a posteriori (MAP) is proposed as a post-processing tool at the decoder side to recover the lost information of video sequences after transmitting over the communication channels. In order to protect the edges, discrimination analysis is used to detect edges. So, edge threshold T for the Huber function of MRF is adaptively obtained, according to the edge or smooth area that the pixel belongs to. In order to eliminate the blocking artifacts, a slope k is also introduced to the linear part of the Huber function. Simulation results show that the proposed method can recover image with the higher quality, comparing to the existing error concealment methods.

SURF feature detection method used in object tracking

Traditional Mean-shift tracking algorithm cannot adjust the tracking windows according to the scale and orientation change of the object during tracking and get accurate localization. This paper combines SURF feature detection with the Mean-shift tracking, which matches the SURF feature in target of current and previous frames, calculate their orientation and proportion of scale to realize a scale and orientation changing tracking algorithm. The algorithm builds a model to describe the motion of target and forecast the location of center, which will get a better initial point and reduce iterations. The experiment results show that the proposed algorithm can disposal the scale and orientation change of target and reduce iterations.

Error concealment based on adaptive MRF-MAP framework

Error concealment is a post-processing tool at the decoder side to recover the lost information of video sequences after transmitting over the noisy communication channels. An adaptive error concealment algorithm based on Markov Random Field (MRF) – Maximum a Posteriori (MAP) framework is proposed. Firstly, Discrimination Analysis is used to detect edges. So, edge threshold T for the Huber function of MRF is adaptively obtained, according to the edge or non-edge area that current pixel belongs to. Then, in order to eliminate the blocking artifacts, a slope k is also introduced to the linear part of the Huber function. Simulation results show that the proposed algorithm can recover images with the higher quality, comparing to the existing algorithms.

A Biased Vector Field Convolution External Force for Snakes

The vector field convolution (VFC) is an effective external force for active contour models. However, it always comes across premature convergence in extracting complex geometries, especially narrow and deep concavity when the initial contour is set outside of the object boundary. In this letter, a biased vector field convolution (BVFC) external force is proposed. In BVFC, an indicator function with respect to the contour and a narrow band are introduced to biasedly utilize the edges gradient information of a concave region. On the other hand, a feature map which better describes the principal curvatures and equally emphasizes both corners and edges is also introduced. Experimental results demonstrate that the BVFC snake improves the performance in extracting object boundary and shows the ability to converge to concavity compared with several state-of-art active contour models.

Sample selection method in supervised learning based on adaptive estimated threshold

Machine learning has been used in many areas, such as object detection, pattern recognition, and data dining. Most machine learning algorithms require a high-quality training set. The performance of machine learning would be improved when the informative samples are selected for training. This paper proposes a straightforward definition of boundary samples and provides an effective method to estimate the threshold. The proposed procedure evaluates the weighted average of distance to estimated thresholds for each sample and then selects the boundary samples from initial data. The experimental results show that detectors trained by the selected data have both a higher detection rate and a lower false positive rate compared to the one using training samples which are selected randomly.

Regularized fuzzy clustering for fast image segmentation

Fuzzy clustering is a popular method for image segmentation and various of models based on fuzzy clustering are proposed. However, many methods suffer from the slow convergence and sensitivity to noise and parameters. In this letter, a novel fuzzy clustering method for image segmentation is proposed to solve these problems. A kernel which incorporates the local spatial information is proposed to regularize the membership partition matrix, the convolution operation between the proposed kernel and membership partition matrix greatly decreases the computational complexity. Because of the proposed kernel, the local neighbor information can be flexibly used, which makes the proposed algorithm robust to noise. Furthermore, the proposed algorithm does not depend on the preprocessing and empirically adjusted parameters any more. Experimental results show that the proposed algorithm is robust to noise, very fast and efficient.

Adaptive Blocking Artifacts Reduction Algorithm Based on DCT-Domain

Block-based DCT compression methods usually result in discontinuities called blocking artifacts at the boundaries of blocks due to the coarse quantization of the coefficients. In order to save more computations, an algorithm directly based on DCT-domain is proposed. It reduces the blocking artifacts as much as possible and preserve edge adequately as well. The algorithm presents a specific and convenient edge detection criterion. For the non-edge blocks, one-dimension DCT is used on each row of two adjacent blocks and the shifted block, and then the transform coefficients of the shifted block are replaced by adaptively weighted average of three blocks coefficients. For the edge blocks, Sigma filter is used to smooth the block boundaries. Simulation results show that the proposed reducing blocking artifacts algorithm outperforms the existing algorithms