An operator of gradient consistency for multimodal image registration

Multimodal image registration is one of the most important and difficult problems in the image registration research field. Due to the nonlinear change in the image gray scale between different modalities, the measure to determine the similarity between the image areas or image features is the main difficulty. By analyzing the characteristics of a multimodal image gradient, we propose the gradient consistency operator based on the norm-weighted angle between gradient vectors to improve the region-based image registration algorithm, and apply it to the registration between medium-wave infrared and visible images.