In-Focus Imaging by Mosaicking and Super-Resolution

Although an all-in-focus image can be generated by mosaicking all the best focused regions of a sequence of acquired images taken under different focal settings, an even better-resolved result can be obtained with those discarded less-focused regions which actually contain more useful information. This paper presents a method to estimate an all-in-focus and super-resolved image of a scene from a sequence of such images. The proposed method applies mosaicking and super-resolution techniques simultaneously to iteratively reconstruct a better image by blind trellis searching. Partitioning images into small blocks makes the new method parallel and fast over blocks. With our regularized approach, a high resolution image can be efficiently estimated both in the underdetermined case and in the case with registration errors.