Multispectral autofocus optimization with a coarse-to-fine search and the Pearson correlation coefficient

In multispectral imaging systems incorporating multiple illumination sources, conventional autofocus algorithms often suffer from poor efficiency and are prone to becoming trapped in local extrema. To address these challenges, we introduce a hybrid coarse-to-fine focus search algorithm enhanced by the Pearson correlation coefficient (PCHC-FocusSearch). The proposed algorithm integrates the global search algorithm with hill-climbing, leveraging Pearson correlation coefficients to identify the optimal sequence of wavelength channels for precise focus. An experimental platform featuring programmable LED arrays, an industrial camera, and a motorized zoom lens was assembled to quantify spectral correlations and validate the algorithm's performance. Compared to golden section search, Fibonacci search, and the HC-FocusSearch variant, PCHC-FocusSearch reduces the average number of focus steps by 40%. It also decreases total focusing time by 45% and consistently converges to the true focal plane across all spectral bands. The results show that the algorithm accelerates the autofocus process while maintaining high accuracy under complex and variable illumination conditions. Consequently, it provides a robust solution for real-time multispectral imaging applications.