Efficient Multi-Focus Image Fusion via Depthmap

Image fusion is a set of operation which is done by combination of two or more input images and the result is a better image with more useful information, in a specific application direction. Images can be taken from the same source with different parameters or from completely different sources. In this paper a novel multi focus image fusion method based on depth map and focused regions is proposed. The combination of depth map with segmented focused from de-focused areas by means of deep convolutional neural networks will improve the fusion performance. In the presented research, the input images are passed through a deep neural networks and then merged with the images obtained from the depth map in accordance with the strategy of image fusion. Our approach can yield much fewer ringing results. Comparative experimental results based on various real images demonstrate that our approach outperforms the state of the art methods.