Paper Details

ABSTRACT

Image fusion is one of the emerging topics in image processing due to its applications in computer vision, military services, medical imaging, remote sensing and so forth. The cameras today have limited depth-of-field. So, in multi-focus image fusion, the images that have different focus areas are merged to produce the all in focus image. In this research work, a block based spatial image fusion algorithm is presented. In block based algorithm, finding a suitable block size is a problem, because it may vary from one image to another. Here, an effective quad tree structure is presented that decomposes the image into blocks only if it is required to do so. If one of the blocks of source images is fully focused and other is fully out-of-focus then fully focused block is copied to the fused image, otherwise the block pair is recursively divided. Also, the level, in quad tree structure, upto which the decomposition takes place is more in proposed algorithm. The algorithm is robust as it can work with any focus measures such as variance, EOMG, spatial frequency, etc. From the experiments carried out, proposed algorithm has comparatively better results of entropy, MSE, PSNR and edge based similarity metrics.