In most of the classification methods for images, only the effective information is utilized, which is extracted from the original image using a priori knowledge. In contrast, this study intends to estimate the qualit...
详细信息
In most of the classification methods for images, only the effective information is utilized, which is extracted from the original image using a priori knowledge. In contrast, this study intends to estimate the quality of the product group from the nonuniform texture image, without using a priori knowledge and using only the result of decision by the expert as supervisor data. A neural network (NN) is used for classification. It is shown first that the classification based on the principal components, which is a linear classification procedure, is difficult. Then three learning methods are considered. In the first method, the small regions of the image are input and the frequency filter is formed by learning in the NN. In the second method, the wavelet component is used as the input. In the third method, the wavelet component is used as the input, and the learning is selectively self-inhibited according to the intermediate response to the small region. In an experiment, it is shown that the first two methods require a long time for learning, although the classes can be formed, and the third method is effective.
A new method for constructing locally supported radial wavelet frame or basis, which is different from the multiresolution analysis, is proposed. A continuously differentiable radial function with a local, support is ...
详细信息
A new method for constructing locally supported radial wavelet frame or basis, which is different from the multiresolution analysis, is proposed. A continuously differentiable radial function with a local, support is chosen at first. Then a radial wavelet is obtained by the first and second derivatives of the radial function. If the radial function is both locally supported and infinitely differentiable, so is the radial wavelet. It is shown that the radial wavelet is a multidimensional dyadic one. I. Daubechies’ wavelet frame Theorem is extended from one dimension ton dimensions. It is proven that the family generated by dilations and translations from a single radial wavelet can constitute a frame inL 2(? n ). Consequently, it is concluded that the radial wavelet family generated by dilations and translations combined with their linear combination can constitute an orthonormal basis inL 2(? n ). Finally, an example of the radial wavelet, which is inseparable and with a local support and infinitely high regularity, is given based on the framework given here. As an application, a class of wavelet network for image denoising is designed, and an underrelaxation iterative fastlearning algorithm with varied learning rate is given as well.
Proposes a hybrid learning algorithm of RBF neural networks. The number of hidden neurons is decided by a network growth technique. A membership function is introduced into training center vectors of Gaussian function...
详细信息
Proposes a hybrid learning algorithm of RBF neural networks. The number of hidden neurons is decided by a network growth technique. A membership function is introduced into training center vectors of Gaussian functions. The reciprocal of the fuzzy factor, which increases during iteration, is considered as the temperature in simulated annealing. This algorithm can not only effectively overcome initial weight sensitivity problems and the dead-node problem of the c-means clustering algorithm, but also dynamically determines the hidden neurons. Experimental results show that the algorithm proposed in the paper is effective.
A novel fast motion estimation method based on an improved genetic algorithm is presented, in which both objective search and random search derived from genetic mutation are used for searching the global optimum and a...
详细信息
A novel fast motion estimation method based on an improved genetic algorithm is presented, in which both objective search and random search derived from genetic mutation are used for searching the global optimum and a threshold selection operator is applied to speeding up the estimation. The selection of initial population based on the coherence between neighboring macroblocks also improves the performance of search. Experimental results demonstrate that this method has very similar performance to that of FS, but just slightly slower than 3SS and 2DL. The inherent robustness and high parallelism enable it to be suitable for VLSI implementation of video encoders.
A new neural network model for unsupervised pattern classification, which is known as generalized entropy mapping (GEM) , is presented. The framework, characteristics and performance of generalized information en-trop...
详细信息
A new neural network model for unsupervised pattern classification, which is known as generalized entropy mapping (GEM) , is presented. The framework, characteristics and performance of generalized information en-tropre neural network are discussed. The GEM can be used for image segmentation in computer vision system. The global optrmization net based on generalized entropy measure is given. The experimental results show that the performance of the GCM net is efficient in low-level visual information processing.
Robotic manipulators in contemporary work-cells are often incapable of solving even simple pick-and-place operations. More specifically, most systems require each object to be supplied in the same and pre-defined way....
详细信息
Robotic manipulators in contemporary work-cells are often incapable of solving even simple pick-and-place operations. More specifically, most systems require each object to be supplied in the same and pre-defined way. Introducing regrasping to the production cycle relaxes the constraints imposed on supply mechanisms. Regrasping describes the operation that must be performed whenever an object's pick-up grasp is incompatible with its put-down grasp. The paper presents a novel approach to solve the regrasp problem for robots equipped with parallel-jaw end-effecters. The proposed method first evaluates the object's possible grasps and placements. These are subsequently combined to form grasp-placement-grasp triples. A regrasp sequence leading from the pick-up to the put-down grasp is generated by searching through the resulting space of compatible grasp-placement-grasp triples. The algorithm takes kinematic and geometric constraints of both the manipulator and the objects into account. Computational concerns are addressed by subdividing the calculation into an off-line and a fast online phase.
A novel type of connection structure between visual cortex and thalamus was proposed. In which included feedback connections from cortical level to thalamic level. Their derivation and biophysical interpretation are p...
详细信息
A novel type of connection structure between visual cortex and thalamus was proposed. In which included feedback connections from cortical level to thalamic level. Their derivation and biophysical interpretation are presented, along with a stability analysis of their dynamics by deriving a global Liapunov functional for so-called CTM system. The system proposed is suitable for computerized digital processing of gray-level images for enhancement.
A class of 2D compactly supported radial wavelet basis with infinitely high regularity is constructed in this *** corresponding wavelet network for image denoising is *** from tensor product multidimensional wavelet,t...
详细信息
A class of 2D compactly supported radial wavelet basis with infinitely high regularity is constructed in this *** corresponding wavelet network for image denoising is *** from tensor product multidimensional wavelet,this one is inseparable,and the characteristics of local support and infinitely high regularity are coexisted instead of conflicting in usual orthogonal wavelet *** underrelaxation iterative fast-learning algorithm with varied learning ratio is *** the computational complexity of algorithm is O(N ) for N×N images,whereas is O(N log N) using Mallat's fast wavelet *** learning algorithm can be implemented using less storage and computation than the orthogonal least squares learning algorithm in radial basis function *** test results show that the wavelet network is suited to remove either Gaussian noise or real noises.
The Volterra operations are an effective way for the exploitations of higher order statistical redundancies of *** present a design method for adaptive quadratic Volterra filters with application to edge detection of ...
详细信息
The Volterra operations are an effective way for the exploitations of higher order statistical redundancies of *** present a design method for adaptive quadratic Volterra filters with application to edge detection of *** filter is expressed as a linear combination of generalized Teager basis filters. The expression coefficients are determined by minimizing the mean square error(MSE).And some simulations are given in this article.
In this study we propose a novel type of connection structure between visual cortex and LGN in thalamus,including feedback connections from cortical level to thalamic *** derivation and biophysical interpretation are ...
详细信息
In this study we propose a novel type of connection structure between visual cortex and LGN in thalamus,including feedback connections from cortical level to thalamic *** derivation and biophysical interpretation are presented,along with a stability analysis of their dynamics by deriving a global Liapunov functional for so-called CTM *** system proposed is suitable for computerized digital processing of gray-level images for enhancement.
暂无评论