A Non-rigid Image Registration Method Based on Improved Optical Flow Model

Abstract

The invention discloses an improved optical flow field model algorithm. According to the invention, the local mutual information method is adopted for solving an initial displacement field between a reference image and a floating image. The Gaussian pyramid image layers of the reference image and the floating image is constructed. In a top image of a pyramid, the initial displacement field is usedfor correcting the floating image layer, and the energy function is minimized. After that, a motion displacement field between the reference image layer and the floating image layer is solved. The last layer of the motion displacement field is used as the initial displacement field of the next layer, and the motion displacement field of the next layer is calculated by utilizing an optical flow equation. The obtained motion displacement field is used as a final motion displacement field until an original image layer is obtained. The floating image is corrected according to the final motion displacement field, and then a registration image is obtained. According to the invention, the estimation precision of the optical flow model on the large displacement deformation of a non-rigid image can be improved. The over-smoothing phenomenon during the optical flow estimation process is avoided. The accuracy of optical flow estimation is improved. Meanwhile, image edge characteristics are kept.The algorithm is applicable in the fields of object motion estimation, target recognition, video compression and the like.

Description

technical field [0001] The invention belongs to the field of computer vision, and more specifically relates to a non-rigid image registration method based on an improved optical flow field model. Background technique [0002] Optical flow refers to the instantaneous velocity of the pixel movement of a space moving object on the observation imaging plane. It is to find the previous frame and the current frame through the change of pixels in the image sequence in the time domain and the correlation between adjacent frames. A method to calculate the motion information of objects between adjacent frames by the corresponding relationship between them. Optical flow not only contains a large amount of motion information of the target object, reflecting the change of the image, but also contains information such as the three-dimensional structure of the target object, and is widely used in motion estimation, target recognition, video detection, image registration and other fields. ...

AI Technical Summary

Problems solved by technology

However, these two assumptions are sometimes not satisfied for non-rigid deformation

For example, when the illumination of the image changes unevenly, the error of the optical flow displacement field estimated by brightness conservation is large, and the floating image cannot be well registered; at the same time, when the object undergoes a large deformation, the distance between the two images The motion displacement field may be larger

At this time, according to the assumption of global smooth consistency, by minimizing the energy function, a reliable displacement field cannot be obtained, which will greatly reduce the accuracy of the registration

On the other hand, a strong assumption of global smooth consistency will also cause the diffusion of optical flow to appear over-smoothing at the edge of the image, which is not conducive to maintaining the edge information of the image. Therefore, where the gradient information of the image changes significantly, the estimated The motion displacement field is prone to large deviations

Images