Fisher classifier and fuzzy logic based multi-focus image fusion
Jamal Saeedi, K. Faez
Dec 28, 2009
Citations
16
Citations
Journal
2009 IEEE International Conference on Intelligent Computing and Intelligent Systems
Full text
Semantic Scholar
Key Takeaway Key Takeaway: This paper proposes a new method for spatially registered multi-focus image fusion using a two-class fisher classifier and fuzzy logic, providing improved subjective and objective results compared to previous methods.
Abstract
In this paper, we propose a new method of spatially registered multi-focus images fusion. Image fusion based on wavelet transform is the most commonly fusion method, which fuses the source images information in wavelet domain according to some fusion rules. We formulate image fusion process as a two class problem: in focus and out of focus classes. Two-class fisher classifier is used for this purpose and six dimensional feature vectors, which is obtained via dual-tree discrete wavelet transform sub-bands are used for training classifier. We use classifier output as a decision map for selecting wavelet coefficients between two images in the different directions and level of decomposition, equally. Also there is an uncertainty about selecting wavelet coefficients in the smooth regions of two images, which causes some misclassified regions. In order to solve this uncertainty and integrate as much information of each source image as possible into the fused image, we propose an algorithm based on fuzzy logic, which combines output of three fusion rules. This new method provides improved subjective and objectives results compared to the previous fusion methods.