Research On Remote Sensing Images Fusion Based On Multi-scale Geometric Analysis And Classification Methods

In this paper, an algorithm of remote sensing object fusion is proposed based onboth the multi-scale geometric analysis and a method that use the Fractal BrownMotion to calculate the fractal dimension, and an classification algorithm is draggedbased on the fusion that between the spectral classifier and the texture classifier whichbased on the fractal dimension used in the texture extraction.Because that the multi-scale geometric analysis have a good performance ingetting a higher image resolution and keeping more the spectral information of animage, and also have better result in the correlation, average difference and thestandard deviation of the image. And the multi-scale geometric analysis is getting moreand more popular in the image analysis. The fractal dimension can reflect the imagetexture information well as a measure of the degree of irregular of the image, so thefractal dimension is also used widely in the image processing. In this paper, themulti-scale geometric analysis is used together with the fractal dimension. Firstly, usethe multi-scale geometric analysis to decompose the images into low frequencycoefficients and high frequency coefficients, and then the Fractal Brown Motion isused to calculate the fractal dimension of the two images that are used to fusion. Thenchoose the high frequency coefficients of the bigger dimension. Lastly, the informationquantity and the average gradient of the image are used to evaluate the algorithm.In the classification algorithm, an classification algorithm is dragged based on thefusion that between the spectral classifier and the texture classifier that based on thefractal dimension used in the texture extraction. Compare the algorithm proposed byHe Longbing, here we use the fractal dimension to extract the texture feature of theimage. And some experiments are done in the last to inspect the algorithmperformance.