Multimodal Image Matching Based on Block Phase Consistency and Orientation Index Map

Due to the differences in imaging mechanisms and nonlinear radiation distortion (NRD) between images of different modalities, multimodal image matching methods encounter challenges in matching performance. To address this issue, we propose a novel method called Block Phase Consistency Image Feature Transform (BPCIF). In the feature detection stage, the BPCIF method employs block-based phase consistency maps for feature point detection, which improves the distribution of detected feature points and yields a more uniform set of highly repeatable feature points. In the feature description stage, the BPCIF method utilizes newly proposed Orientation Index Map (OIM) based on log-Gabor filtering for feature point description, which significantly reduces the impact of NRD and significant contrast differences between multimodal images. By leveraging information from different filter orientations during the OIM construction process, the matching accuracy is significantly improved. Additionally, we introduce a new descriptor construction method based on OIM to achieve rotational invariance while ensuring efficient descriptor computation. Experimental results demonstrate that BPCIF outperforms the state-of-the-art methods in terms of the quantity and accuracy in multimodal image matching.