Font Size: a A A

Research On Multi-focus Image Fusion Algorithms

Posted on:2013-02-18Degree:MasterType:Thesis
Country:ChinaCandidate:X X MaFull Text:PDF
GTID:2218330371464537Subject:Detection Technology and Automation
Abstract/Summary:PDF Full Text Request
The essence of multi-focus image fusion is to extract the clear area from every source image, and integrate these clear areas to produce the fused images including all the scenes. This paper mainly studies the pixel-level multi-focus image fusion algorithms. Some solutions to the problems caused by existing image fusion algorithms are introduced in this paper.In space domain, block-based assimilation of spatial frequency multi-focus image fusion algorithm is proposed. Spatial frequency is the clarity of an image. It reflects the activity of the word block. The spatial frequency of clear block is above the ambiguous block. So we can use the spatial frequency in the algorithm and thus accurately extract the clear areas from source images and fuse them in one image. Using the method of block- assimilation is to reduce the error and smooth the critical region. Simulation results show that this algorithm can accurately extract the clear areas, avoiding the phenomenon of "ringing" and ghosting and improves the quality of image fusion.From the complementary of space domain and transform domain, PCA-based Laplacian pyramid transform on multi-focus image fusion algorithm is proposed. Because the traditional algorithm in space domain lack of the detail expression, using the method of Laplacian pyramid transform to decompose the source images. The algorithm of PCA can retain the key information of an image, so that the top-level image of pyramid images uses the PCA algorithm. Average gradient reflects the small contrast and texture features, so that the other layers of images use the average gradient algorithm. Simulation results show that this algorithm improves image clarity and achieve the desired results.In transform domain, wavelet-based Laplacian pyramid multi-focus image fusion algorithm is proposed. Pyramid decomposition algorithm based on non-directional. But it's date is redundancy and correlation between layers. Mean the time, the wavelet transform algorithm has no redundant information in different decomposition layers, but can't capture all of the orientation information. As the result, get the two algorithms together. Firstly, decompose the image by wavelet transform and get the high-frequency information and low-frequency information of the image. After that, select the low-frequency coefficient by pyramid algorithm and select the high-frequency coefficient by the max value of the clarity. Simulation results show that the algorithm avoids the ghosting and improve the quality of the image fusion.
Keywords/Search Tags:Image fusion, Multi-focus images, Spatial frequency, PCA, Laplacian pyramid, Wavelet transform
PDF Full Text Request
Related items