Construction Method Of Directional Filter Bank And Its Application In Image Fusion

The rich directional features and texture details have important influence on the extraction of image edge information in image processing,which is very important for image fusion and further analysis.The traditional image fusion method based on contrast pyramid decomposition can provide the physical meaning of contrast for the image,but there is no emphasis on direction detection.Discrete wavelet transform is widely applied to image fusion,but the direction information it provides is limited.Therefore,we are committed to seeking a tool which can efficiently extract the orientation information of image.Considered the characteristic of the directional filter bank,it can effectively extract the directional edge features of the image.The directional filter bank with superior performance is constructed,and an image fusion method based on directional contrast pyramid decomposition is proposed in this paper.The major research works of this paper are as follows:According to the design theory of directional filter banks,the directional filter banks are constructed in this paper.The core of the directional filter bank in the Contourlet transform is the quincunx filter.The division of 2l directional sub-bands is obtained by rotation and down-sampling for the image.In order to solve the problem that the above method has the disadvantages of rotation and down-sampling,two kinds of all phase directional filter banks with different structures are constructed here.The one is multi-level structure,and the other is iterative structure.A pair of all phase fan filters are constructed from the frequency spectrum of the fan filters by designing the column response.Then,a pair of all phase quadrant filters are obtained by rotating the all phase fan filter templates at 45 degree.In this way,the structure of the first two levels is implemented by cascading of all phase fan filters and all phase quadrant filters.Since the third level,there are two construction structures that multilevel structure and iterative structure.The filters at all levels are still designed by the column response,which is the former structure.The latter is obtained by the iteration of the fan filters and image rotation operation.An image fusion method based on directional contrast pyramid decomposition is proposed in this paper.First the contrast pyramid decomposition of multi-focus images is performed.Then,the high frequency sub-images with different scales and different directions are obtained for each high frequency image by using the directional filter bank.According to the characteristics of different frequency domain,the decomposed sub-images are fused,and the fusion rule is that the weighted average value is selected for low frequency sub-band coefficients,while for high frequency sub-band coefficients,the coefficients with larger absolute value are employed.Finally,the performance of the result image is analyzed and evaluated by visual experience and objective indices.The experimental result shows that the proposed method can provide rich direction information while maintaining the meaning of contrast.The fused image obtained by the proposed method not only has better visual effect,but also has higher definition and stronger spatial resolution.