Catálogo de publicaciones - libros

This study aimed to develop an advance precision three-dimensional (3-D) image segmentation algorithm to enhance the blurred edges clearly and then introduce the result onto the intensity modulated radiotherapy (IMRT) for tumor target volume definition. This will achieve what physicians usually demand that tumor doses escalation characteristics of IMRT. A proposed algorithm flowchart designed for this precision 3-D treatment targeting was introduced in this paper. Different medical images were used to test the validity of the proposed method. The 3-D wavelet based targeting preprocessing segmentation allows physicians to improve the traditional treatments or IMRT much more accurately and effectively. This will play an important role in image-guided radiotherapy (IGRT) and many other medical applications in the future.

After the discrete stationary wavelet transform (DSWT) combined with the generalized cross validation (GCV) for an image, the noise in the image is directly reduced in the high frequency sub-bands, which are at the high- resolution levels. Then the local contrast of the image is enhanced by combining de-noising method with in-complete Beta transform (IBT) in the high frequency sub-bands, which are at the low-resolution levels. In order to enhance the global contrast for the image, the low frequency sub-band image is also processed by combining the IBT and the simulated annealing algorithm (SA). The IBT is used to obtain the non-linear gray transform curve. The transform parameters are determined by the SA so as to obtain the optimal non-linear gray transform parameters. In order to reduce the large computational requirements of traditional contrast enhancement algorithms, a new criterion is proposed with the gray level histogram. The contrast type for an original image is determined by employing the new criterion. The gray transform parameters space is respectively given according to different contrast types, which greatly shrinks gray transform parameters space. Finally, the quality of the enhanced image is evaluated by a new overall objective criterion. Experimental results demonstrate that the new algorithm can greatly improve the global and local contrast for an image while efficiently reducing gauss white noise (GWN) in the image. The new algorithm performs better than the histogram equalization (HE) algorithm, un-sharpened mask algorithm (USM), Tubbs’s algorithm [2], Gong’s algorithm [3] and Wu’s algorithm [4].

On the basis of Markov Random Field (MRF), which uses context information, in this paper, a robust image segmentation method is proposed. The relationship between observed pixel intensities and distance between pixels are introduced to the traditional neighbourhood potential function, which described the probability of pixels being classified into one class. To perform an unsupervised segmentation, the Bayes Information Criterion (BIC) is used to determine the class number. The K-means is employed to initialise the classification and calculate the mean values and variances of the classes. The segmentation is transformed to maximize a posteriori (MAP) procedure. Then, the Iterative Conditional Model (ICM) is employed to solve the MAP problem. In the experiments, the proposed method is adopted with K-means, traditional Expectation-Maximization (EM) and MRF image segmentation techniques, for noisy image segmentation applying on synthetic and real images. The experiment results and the histogram of signal to noise ratio (SNR)-miss classification ratio (MCR) showed that the proposed algorithm is the better choice.

Blur identification is usually necessary in image restoration. In this paper, a novel blur identification algorithm based on Support Vector Machines (SVM) is proposed. In this method, blur identification is considered as a multi-classification problem. First, Sobel operator and local variance are used to extract feature vectors that contain information about the Point Spread Functions (PSF). Then SVM is used to classify these feature vectors. The acquired mapping between the vectors and corresponding blur parameter provides the identification of the blur. Meanwhile, extension of this method to blind super-resolution image restoration is achieved. After blur identification, a super-resolution image is reconstructed from several low-resolution images obtained by different foci. Simulation results demonstrate the feasibility and validity of the method.

Super-resolution image reconstruction estimates a high-resolution image from a sequence of low-resolution, aliased images. The estimation is an inverse problem and is known to be ill-conditioned, in the sense that small errors in the observed images can cause large changes in the reconstruction. The paper discusses application of existing regularization techniques to super-resolution as an intelligent means of stabilizing the reconstruction process. Some most common approaches are reviewed and experimental results for iterative reconstruction are presented.

Inspired by studies of cognitive psychology, we proposed a new dynamic similarity kernel for visual recognition. This kernel has great consistency with human visual similarity judgement by incorporating the perceptual distance function. Moreover, this kernel can be seen as an extension of Gaussian kernel, and therefore can deal with nonlinear variations well like the traditional kernels. Experimental results on natural image classification and face recognition show its superior performance compared to other kernels.

In this paper, we present an extended boids model for simulating the aggregate moving of fish schools in a complex environment. Three behavior rules are added to the extended boids model: following a feed; avoiding obstacle; avoiding enemy boids. The moving vector is a linear combination of every behavior rule vector, and the coefficients should be optimized. We also proposed a genetic algorithm to optimize the coefficients. Experimental results show that by using the GA-based optimization, the aggregate motions of fish schools become more realistic and similar to behaviors of real fish world.

Automated segmentation of MR images is a difficult problem due to complexity of the images. In this paper, we proposed a new method based on independent component analysis (ICA) for segmentation of MR images. We first extract thee independent components from the T1-weighted, T2-weighted and PD images by using ICA and then the extracted independent components are used for segmentation of MR images. Since ICA can enhance the local features, the MR images can be transformed to contrast-enhanced images by ICA. The effectiveness of the ICA-based method has been demonstrated.

We consider the problem of partitioning a data set of data objects into homogeneous subsets (that is, data objects in the same subset should be similar to each other), such that each subset is of approximately the same size. This problem has applications wherever a population has to be distributed among a limited number of resources and the workload for each resource shall be balanced. We modify an existing clustering algorithm in this respect, present some empirical evaluation and discuss the results.

The Fisher kernel, which refers to the inner product in the feature space of the Fisher score, has been known to be a successful tool for feature extraction using a probabilistic model. If an appropriate probabilistic model for given data is known, the Fisher kernel provides a discriminative classifier such as support vector machines with good generalization. However, if the distribution is unknown, it is difficult to obtain an appropriate Fisher kernel. In this paper, we propose a new nonparametric Fisher-like kernel derived from fuzzy clustering instead of a probabilistic model, noting that fuzzy clustering methods such as a family of fuzzy -means are highly related to probabilistic models, e.g., entropy-based fuzzy -means and a Gaussian mixture distribution model. The proposed kernel is derived from observing the last relationship. Numerical examples show the effectiveness of the proposed method.