Guided image filtering in shape-from-focus: A comparative analysis

Abstract

Mostly, shape from focus (SFF) methods do not consider any prior to extend the accuracy of the depth map. Ultimately, even the improved depth map might lack the accurate structure of the object. While reviewing the guided filters, it has been observed that SFF has not been considered as an application. In this study, we not only suggest to apply guided filtering for depth enhancement but also provide a comparative analysis of recently proposed guided filters for SFF framework in a systematic way. In addition, a set of potential guidance maps has been suggested and the performance of these guidance maps has been evaluated. The improved performance of guided filters has been ranked against the depth maps of synthetic and real image sequences where the corresponding scenes have diverse range of geometrical complexities. It has been observed that guided image filtering is effective in improving the initial depth maps in SFF.