Entity

Time filter

Source Type


Cui Z.,R and nter for Modern Information Technology Application in Enterprise | Cui Z.,Soochow University of China | Zhang G.,Soochow University of China
Journal of Software | Year: 2010

Medical image classification as an important research topic both in image processing and biomedical engineering. The ridgelet transform has good directional selective ability to locally and sparsely in representing the image compared with the traditional wavelet transform. This paper proposes a novel classification model for medical image, which is using ridgelet transform and dynamic fuzzy theory. Firstly, the image was decomposed by digital ridgelet transform to obtain the approximation coefficients and detailed coefficients in different sub-bands with directional parameters. Then the dynamic fuzzy theory was applied to construct a membership function to calculate coefficients from each sub-bands respectively, and a weight of sub-bands degree was adjust by precision requirement. At last similarity degrees are calculated by coefficients degree and weight. Medical images were classified by the result sort order of the degrees effectively. © 2010 ACADEMY PUBLISHER. Source


Zhang G.,Soochow University of China | Cui Z.,Soochow University of China | Cui Z.,R and nter for Modern Information Technology Application in Enterprise
Key Engineering Materials | Year: 2011

Graph cuts as an increasingly important tool for solving a number of energy minimization problems in computer vision and other fields, meanwhile beamlet transform as time-frequency and multiresolution analysis tool is often used in the domain of image processing, especially for image fusion. By analyzing the characters of DSA medical image, this paper proposes a novel DSA image fusion method which is combining beamlet transform and graph cuts theory. Firstly, the image was decomposed by beamlet transform to obtain the different subbands coefficients. Then an energy function based on graph cuts theory was constructed to adjust the weight of these coefficients to obtain an optimum fusion object. At last, an inverse of the beamlet transform reconstruct a synthesized DSA image which could contain more integrated accurate detail information of blood vessels. By contrast, the efficiency of our method is better than other traditional fusion methods. Source


Gu L.,R and nter for Modern Information Technology Application in Enterprise | Gu L.,Nanjing University of Posts and Telecommunications
Proceedings - 2012 International Conference on Computer Science and Service System, CSSS 2012 | Year: 2012

In this paper, a nearest neighbor rule is applied to the clustering method based on one-class support vector machines. Although the traditional clustering method inspired the k-means clustering employs the kernel-based one-class support vector machines in improving the clustering performance, it forms the coarse decision boundaries. So this paper uses a nearest neighbor rule to establishing the better decision boundaries. Experimental results show that the novel clustering algorithm can increase the clustering accuracies according to a nearest neighbor rule. © 2012 IEEE. Source


Gu L.,R and nter for Modern Information Technology Application in Enterprise | Gu L.,Nanjing University of Posts and Telecommunications
2012 IEEE 5th International Conference on Advanced Computational Intelligence, ICACI 2012 | Year: 2012

Semi-supervised clustering takes advantage of a small amount of labeled data to bring a great benefit to the clustering of unlabeled data. Based on a locality sensitive k-means clustering method, this paper presents two novel semi-supervised clustering algorithms inspired by the semi-supervised variants of the k-means clustering by seeding. To investigate the effectiveness of our approaches, experiments are done on one artificial dataset and three real datasets. Experimental results show that two proposed methods can improve the clustering performance significantly compared to other unsupervised and semi-supervised clustering algorithms. © 2012 IEEE. Source


Gu L.,R and nter for Modern Information Technology Application in Enterprise | Gu L.,Nanjing University of Posts and Telecommunications
2012 IEEE 5th International Conference on Advanced Computational Intelligence, ICACI 2012 | Year: 2012

The initialization of one clustering method based on one-class support vector machines often employs random samples. This way can lead to the unstable clustering results. In this paper, the k-harmonic means clustering takes the place of this random initialization. To investigate the effectiveness of the novel proposed approach, several experiments are done on one artificial dataset and two real datasets. Experimental results show that our presented method can not only obtain the stable clustering accuracies, but aloes improve the clustering performance significantly compared to other different initialization, such as random initialization and k-means initialization. © 2012 IEEE. Source

Discover hidden collaborations