Multifocus Image Fusion Using PC

In the process of multifocus image fusion, the key issue is to find the focus regions. As it does in the case of square signal whose phase congruency(PC)reaches its maximum at step position. The step position of square signal is similar to sharp pixels at the focus regions of multifocus image, thus we can detect the sharp pixels by using the phase congruency. In addition, the phase congruency is invariant to variations in image illumination and contrast, so the phase congruency is able to detect the focus regions even though the image has been deformed. The phase congruency of the focus measure(FM) which is from the input image in different scales will be better able to gain some of image details. The phase congruency is used as a focus measure and then we address the algorithm of fusing multifocus images based on the phase congruency.The proposed image fusion scheme can be divided into the following steps.First, we compute the value of the FM which is the phase congruency to obtain the FM matrix. Then, the FM matrix is divided into small blocks to obtain the feature matrix so that pixels of focus regions have some relations. Finally, feature matrix is expanded and smoothed by majority filter, and then the fusion image is obtained based on a new fusion rule(FR). The boundary between the focused region and unfocused region definite are rendered by filtering processing. FR is implemented in the spatial domain so that we can get more information from the input image for fusion image.We add Gaussian noise into input images then calculate the FMs of degraded images by using the proposed algorithm. And the experimental results are shown that PC is more robust against noise than other four FMs. Through the experiment, this paper proves that the PC can more accurately calculate the focus regions in the multifocus image than the other four common FMs. Experimental results show that the proposed fusion scheme achieves the fusion performance of the state-of-the-art methods in terms of visual quality and quantitative evaluations.