Depth From Focus in 3-D Measurement: An Overview

The 3-D reconstruction constitutes a fundamental methodology in 3-D measurement, with applications spanning industrial measurement, quality control, autonomous systems, and precision manufacturing. Depth from focus (DFF), as a monocular optical 3-D reconstruction technique, is particularly relevant in these domains due to its cost-effectiveness, configuration simplicity, and compatibility with existing measurement systems. This article provides a thorough review of DFF techniques, focusing on their role in accurate and efficient 3-D measurement. We systematically analyze the principles underlying DFF, along with significant advancements in measurement algorithm performance over two decades. Furthermore, we review the datasets and evaluation methods that are essential for validating the effectiveness of DFF in both camera measurement systems and microscopic measurement systems. Our experimental analysis compares the performance of current deep learning-based and conventional model-based DFF algorithms using publicly available code and measurement benchmarks. The results indicate that deep learning-based algorithms demonstrate superior performance in portable measurement applications, while model-based approaches are more suitable for high-accuracy measurement scenarios such as microscopic metrology. We also highlight the ongoing challenges and emerging opportunities in advancing DFF techniques for the next-generation instrumentation and measurement systems.