Enhancement of Depth Map through Weighted Combination of Guided Image Filters in Shape-From-Focus

Abstract

Estimation of depth map plays a key role in a number of computer vision applications. Shape from focus is a monocular approach that uses image focus as a cue to reconstruct 3D shapes. In the literature, a variety of guided image filters have been proposed to enhance the depth map individually. Among them, some have excess computational time burden, and some others produce unsatisfactory results. This paper proposed a framework for the enhancement of depth map by using a weighted combination of selected guided filters in shape from focus. The optimized weights are obtained using the particle swarm optimization approach, and the subset of best-performing filters is identified through a sequential forward search method. The experimental results have demonstrated that the proposed framework provides considerably improved depth maps.