A progressive optimization method for image alignment of cameras with different focal lengths

Existing techniques are insufficient for directly addressing the challenge of image alignment between long and short focal length cameras in multicamera systems. This inadequacy arises from disparities in lens projection, camera mounting, and imaging posture. To overcome these issues, a novel method that utilizes a coarse-to-precise approach for image alignment is proposed. Initially, the image undergoes transformation to an approximate low resolution, enabling rapid feature extraction and pre-alignment. Subsequently, a dense and uniform set of new feature point pairs is introduced at a higher resolution. A cross-enhanced grid transformation model is then established, facilitating accurate image alignment. To validate the efficacy of this method, simulation experiments were conducted using a single zoom lens, while real-life experiments were performed using a dual camera system comprising long- and short-focus lenses. The experimental results demonstrate that the method proposed in this paper exhibits superior alignment capabilities for local details and enables accurate image alignment of long-and short-focus cameras.