Extended focused imaging in microscopy using structure tensor and guided filtering

Abstract

The limited depth-of-field (DOF) in optical microscopic imaging systems prevents them from capturing an entirely focused images when the imaged scene is thick and covers a wide range of depth. Specimens lying outside the DOF is defocused and thus blurred, hindering the subsequent observation and analysis. Extended focused imaging aims at yielding a single all-in-focus image by combining the in-focus information from a stack of partially focused images. In this paper, we present a new algorithm for digitally extending the low DOF of a microscopic imaging system. The method is based on the two-scale decomposition of an image into a base layer and a detail layer, both of which contain information of different scales. Then, with the structure tensor, which is uniquely advantageous for autofocusing multi-focal planes, as the focus measure, initial weight maps are generated. The guided filtering is then employed to optimize the weighted averaging scheme for the fusion of the base and detail layers. We examine the proposed algorithm under incoherent and coherent imaging modalities, i.e., bright-field microscopy and digital holographic microscopy. Qualitative and quantitative comparisons with the existing approaches are carried out using extensive experimental data. The results show the advantages and superior performance of the proposed scheme even in the presence of coherent noise.