Auto-focusing evaluation functions in digital image system

Abstract

Evaluation function is an important item to measure the quality of images in auto-focusing systems. This paper introduces four categories of evaluation functions which are respectively based on gradient, correlation, statistics and transform. These functions formally differ from each other and their performances are not consistent when dealing with different images. The same function also performs different when the processed images have different characteristics. Five commonly used functions from the four categories are verified and compared under different circumstances in order to validate their performances. The result provides references for variety of selections in the practical application.