| Summary: | 碩士 === 國防大學理工學院 === 電子工程碩士班 === 100 === Because of its widely application, many techniques for multi-focus image fusion have been developed. However, the past fusion methods have their own disadvantages. For example, pyramid algorithm may produce block effect due to the spatial-direction selectivity is not considered in decomposition process and resulted in residue between two different scale and unstable structure. The method of multi-resolution analysis would cause the blurring phenomenon on image. According to the above reasons, a novel image fusion algorithm based on fractal dimension (FD) is proposed in this paper. Expect to get a better fusion image. First, each focus image is decomposed using discrete wavelet transform (DWT) separately. Second, calculate FD of each pixel on transformed images. In light of the Box-Counting, calculate the respective self-narrow ratio for each pixel and the total number of the boxes. Then, use the least squares linear regression to calculate the fractal dimension. Third, the fractal dimension trend selection (FDTS) is proposed to compute the new wavelet subband coefficient. Then, use local trend selected revision to revise the new wavelet subband coefficient. Finally, the generated image is reconstructed from the newest wavelet subband coefficients. Moreover, the reconstructed image can represent more detailed for the obtained scene. Experimental results demonstrate that our scheme performs better than the traditional discrete cosine transform (DCT) and discrete wavelet transform method in both visual perception and quantitative analysis.
|
|---|
