In this paper, we propose a novel technique for blind image restoration and resolution enhancement based on radial basis function (RBF) neural network. The RBF network gives a solution of the regularization problem often seen in function estimation with certain standard smoothness functional used as stabilizers. A RBF network model is designed to represent the observed image. In this model, the number and distribution of the centers (which are set to the pixels of the observed image) are fixed. In addition, network output is set to the observed image pixel gray scale value. The RBF plays a role of point spread function. The technique can also be applied to image resolution enhancement by generating an interpolated image from the low resolution version. Experimental results show that the learning algorithm can effectively estimate the model parameters and the established neural network model has a high fidelity in representing an image. It is believed that the proposed neural network model provides a valuable tool for image restoration and resolution enhancement and holds promises to improve the quality and efficiency of image processing.
In this study, five classifiers, namely quadratic discriminant analysis, linear discriminant analysis, regularlized discriminant analysis, leave-one-out covariance matrix estimate and Killback-Leibler information measure based method are considered for classification of stellar spectra data. Because stellar spectra data sets are severly ill-posed, we first adopt some feature selection method such as principal component analysis to reduce data dimensionality. The input of the classifiers are those selected features, and the cross-validation technique is used to optimize the regularization parameters. Experimental results show that in most cases, regularized classifiers are high classification rates than that of quadratic discriminant analysis, but parameter optimization is time consuming. From experiments of exhaustive searching regularization parameter, it is found that in some cases cross-validation method is not always good in the selection of models.
A new physical model for volume hologram was proposed. Volume hologram was considered as a series coupled Fabry- Perot etalon. The multi beam interference makes band width of volume hologram to be very narrow. The positive feedback of Fabry-Perot provide high diffraction efficiency of volume hologram. Computer simulations based on the new model showed a good consistency with the coupled wave theory and previous experimental results.
A new image nonlinear segmentation method which is based on feedforward multilayer neural network (MLN) is presented in this paper. The example of using proposed MLN technique for cross overlapped chromosome image segmentation is given. In contrast to gray-level threshold technique, the MLN method is based on spatial coordination classification. From the experiments it can be concluded that the MLN in particular shows promise of being a useful method for image nonlinear segmentation.