Construction of fused image with improved depth-of-field based on guided co-occurrence filtering

Abstract With the advent of biomedical imaging systems and the rapid advancements in light microscopy, multi-focus image fusion has provided significant attention in creating a fused image with improved depth-of-field (DOF). The fused image plays an imperative role in different computer vision based applications. Firstly, the proposed algorithm decomposes the source multi-focus images into base layer (BL) and detail layer (DL) using Guided Filter (GF). BL retains large-scale intensity variation and DL is the residual approximation of the source images. Then, a criterion function is developed for in-focus region detection across a set of source multi-focus images. In particular, edge-aware filtering based texture analysis is utilized to detect in-focus saliency maps. These saliency maps are further refined using a guided co-occurrence filter (GCOF) to decide fusion rules. Separate fusion rules for BL fusion and DL fusion are computed adaptively that improve the capability of the proposed fusion process. Finally, using a weighted average fusion (WAF) technique the fusion rules are utilized to construct all-in-focus fused image. The quantitative and qualitative analysis of experimental results for various microscopy stacks and standard photographic multi-focus data sets are discussed to demonstrate the performance of the proposed method.