Color Image Of The Pseudo-color Processing

Pseudo-color processing of the color image, which has a wide range of potential applications, is an active research area of the processing of color image. There are many methods for Pseudo-color processing of the color image, in which the method based on image segmentation is in common use. The important technology of this method is how to segment image efficiently, therefore, this thesis lays a strong emphasis on how to improving the quality of the image segmentation.In this thesis, an approach that connects Gabor filtering with networks is proposed and is used in color image segmentation. Gabor filteing is used in extracting the features of color images, and a neural network is used for the classification of the color image features.First, the effective features are extracted by Gabor filtering. Inspired by the multi-channel operation of the Human Visual System for interpreting texture, research has been focused on a multi-channel approach based on Gabor filtering to the segmentation of color image. Gabor filters have the ability to perform multi-resolution due to its localization both in spatial and spatial frequency domain. Normally the effective width of a filter in the spatial domain and its bandwidth in the spatial-frequency domain are inversely related according the 2D uncertainty principle, therefore, we get the filtered images through a set of Gabor filters arranged as a set of wavelet bases. We have had computer simulating experiments which prove that the multi-channel approaches to image segmentation achieves a good result and make preparations for the follow up.Then, traditional BP neural network and improved BP neural network are used to image segmentation respectively. As a new methodology system, neural network has very strong self-adaptability. Therefore, it has been used widely in many fields, especially in pattern recognition.Traditional BP neural network is a multi-layer feed-forward network that is basedon error back-propagation algorithm and has the widest application. Although traditional BP neural network is successful, it has some disadvantages. For example, its learning convergent velocity is slowly, possibility of converging to a local minimum is high and so on. Therefore, an improved BP neural network is proposed in this thesis and is used in image segmentation. And it is also compared with traditional BP neural network in the aspect of image segmentation. Experimental results show that the improved BP neural network overcomes some disadvantages of the traditional BP neural network effectively.Experimental result show that the proposed approach combining Gabor filtering and neural network performs better in segmentation of color image. The proposed method has stronger self-adaptability and greater segmentation speed.