Fish eye video real-time panoramic correction method based on spherical surface spreading model

Abstract

The invention provides a fish eye video real-time panoramic correction method based on a spherical surface spreading model and belongs to the field of image video processing. The method includes the following steps: fish eye image acquisition, fish eye image pre-processing, fish eye image profile extraction, image geometrical relation converting, re-projection based on fish eye image, and fish eye video roaming. According to the invention, the method includes the following specific steps: after acquiring a fish eye video, firstly resolving the radius of a corresponding "spherical model" of a fish eye image, then performing selection, and restoring each pixel of each frame of fish image to one "spherical model"; then expanding each pixel of the spherical surface to one correction plane according to a corresponding spherical arc; and finally conducting sampling and interpolation on the corrected pixels to obtain a corrected image of the fish eye image. According to the invention, compared with algorithms of same types, the method overcomes the limitation of wide viewing angle and timeliness, addresses unexpandability of the spherical surface, effectively increases image correction precision, achieves fish eye video panoramic expanding, meets real-time applications, has certain robustness for fish eye cameras of different models, and has strong practicality.

Description

technical field [0001] The invention belongs to the field of image and video processing, in particular to a fisheye image correction method based on a spherical expansion model, and realizes real-time panorama expansion of the fisheye video. Background technique [0002] The fisheye lens has the advantages of compact structure, small size, and not easily damaged; at the same time, considering that the omnidirectional vision constructed by the fisheye lens can obtain a true panoramic spherical image, there is no blind spot, and there is no need to consider issues such as image stitching and embedding. Therefore, fisheye lens is one of the most effective methods to construct an all-round vision system, which has high research value and broad potential application prospects. At present, it has a wide range of applications in the fields of video surveillance, video conferencing, virtual reality and robot navigation. However, the images captured by the fisheye camera have very s...

AI Technical Summary

Problems solved by technology

[0005] (1) The fisheye image correction method based on geometric projection is simple, but limited by the wide viewing angle of the fisheye lens, the effective correction area is limited;

[0006] (2) The fisheye image correction method based on camera calibration has high correction accuracy for specific lenses, but it cannot satisfy the distortion correction of multiple fisheye lenses at the same time;

[0007] (3) The fisheye image correction method based on artificial intelligence avoids the limitation of specific lens correction, but the training samples are large, the calculation complexity is high and the storage space is large, so it is not easy to implement in hardware

Images