Image Fusion Based On Guided Filter And Multi-scale Analysis

Image fusion which is developed in image processing technology based on the theory of information fusion technology, is from multiple different sensors or in the different time from the same scene into an image which is to get a more accurate and clear said, making it more suitable for human visual perception and the subsequent image processing. From multiscale geometric analysis, this paper has discussed the image fusion based on multiscale geometric analysis algorithm and its improved method, and apply them successfully in a variety of image fusion, improving the fusion effect. The main work summarized as follows:This paper has improved and innovation something based on the wavelet fusion respectively to multi-focus image fusion, the remote sensing image of multi-spectral image and panchromatic image fusion, and medical image fusion. In the multi-focused image fusion, this paper takes 97 wavelet decomposition having the source image corresponding into the low frequency based on the global contrast of image fusion of significant rule and high frequency based on guide filtering fusion rules.Because the image carries the image of most of the energy in low frequency, shows the outline of image information, thus the fusion rule based on significant to perform well in image contour information.Due to the high frequency information performs the detail of the image characteristics, in order to make the fused image edge character more smooth, joined the orientation filter, making the effect of the fused image better.For multi-spectral image and panchromatic image fusion, this paper adopted a multi-scale fusion method based on IHS and guide filter.As a result of the multispectral image has rich spectral information, but the spatial resolution is not high. Panchromatic images has higher spatial resolution but is not rich in spectral information. Thus the fused images should not only have abundant spectral information, and reaches a certain spatial resolution.In this paper, the method first put the IHS transform into multi-spectral and panchromatic images, then making the I component wavelet decomposition respectively,using guide filter fusion rule in high frequency.The I component of multi-spectral is guide graph, and panchromatic image of I component is input graph.Then making wavelet inverse transformation and IHS inverse transformation, we can get the fusion result.In this paper, the fusion method is a good way to get rich spectral information, the fusion of high spatial resolution image.For the medical image fusion, due to various kinds of medical image, now for medical image fusion is a good way to help the doctor make test condition.Fusion method based on PCA dictionary is proposed in this paper, first of all to CT and MRI image wavelet decomposition, the low frequency using absolute value in large fusion rule, in the high frequency using PCA fusion rule of the dictionary.Because CT image is a good way to perform bone image information, whereas MRI images ability of soft tissue expression is stronger. The proposed approach is a good way to fusion of CT image information about bone information and MRI images of soft tissue information, reaching a better fusion effect.