Multi-Focus Image Fusion based on Gradient Transform

TL;DR

This paper introduces a robust gradient-based multi-focus image fusion method that effectively addresses shift variance and misregistration issues, achieving high success rates in objective evaluations.

Contribution

A novel gradient information-based fusion technique using H-IH transform and majority voting, improving robustness over existing methods.

Findings

Achieved 83.3% success in quantitative metrics.

Outperformed 17 conventional and novel techniques.

Demonstrated robustness against shift variance and misregistration.

Abstract

Multi-focus image fusion is a challenging field of study that aims to provide a completely focused image by integrating focused and un-focused pixels. Most existing methods suffer from shift variance, misregistered images, and data-dependent. In this study, we introduce a novel gradient information-based multi-focus image fusion method that is robust for the aforementioned problems. The proposed method first generates gradient images from original images by using Halftoning-Inverse Halftoning (H-IH) transform. Then, Energy of Gradient (EOG) and Standard Deviation functions are used as the focus measurement on the gradient images to form a fused image. Finally, in order to enhance the fused image a decision fusion approach is applied with the majority voting method. The proposed method is compared with 17 different novel and conventional techniques both visually and objectively. For objective evaluation, 6 different quantitative metrics are used. It is observed that the proposed method is promising according to visual evaluation and 83.3% success is achieved by being first in five out of six metrics according to objective evaluation.