A multifocus image fusion technique based on kernel Fisher discriminant analysis and image block segmentation is proposed. Firstly

the original images are decomposed into image blocks and focus measures of each image block are computed. To achieve the parameters of the trained kernel Fisher discriminant analysis

parts of the original images are chosen as the training exemplars. Then the initial fused image is acquired with the known kernel Fisher discriminant analysis. At last

the final merged image is obtained after the original image blocks

which are located near the border between the focused and blurred areas of the original images

through processing with the redundant wavelet transform. The experimental results show that the proposed method outperforms the conventional image fusion methods. A better tradeoff can be obtained between improving the image fusion quality and reducing the computational cost with the proposed method.