A method for automatic image warping and image stabilization based on linear programming

Abstract

The invention discloses a method for automatically distorting and stabilizing a video image based on linear programming. First, the original video is separated frame by frame to obtain a picture set, and then the estimated affine transformation matrix between two adjacent frames of the original video is used as motion information, extract the straight line segment in each frame, estimate the vanishing point position in the vertical direction and the position of the horizontal line; then build a linear programming model, and use the transformation matrix used to correct each frame to be the parameter to be optimized of the linear programming model, The number of projective transformation matrices equal to the number of video frames is obtained by solving the model; then the original video is transformed frame by frame, and the transformed frames are synthesized to obtain a new video. A series of projective transformation matrices optimized by the invention can simultaneously complete the warping and image stabilization, the processed video has better visual effect, the processing speed is fast, and the obtained result has global optimality.

Description

technical field [0001] The invention relates to video image stabilization and video correction technology, in particular to a method for automatic correction and image stabilization of video images based on linear programming. The method of automatic video correction and image stabilization based on linear programming can quickly and automatically perform image stabilization and correction for video at the same time. Background technique [0002] Due to camera shake, inappropriate jitter is often present in the captured video. Moreover, vertical structures such as buildings are often inclined in the video due to the shooting angle and other reasons, which reduces the visual comfort of the video. Existing video stabilization methods generally restore the original 3D path of the camera through a motion recovery structure (Structure from Motion, SfM), or describe the 2D path according to the position change of feature points or the transformation matrix between adjacent frames...

AI Technical Summary

Problems solved by technology

Some warping methods for a single image use these geometric information for image warping. However, such methods can usually only be modeled as non-convex optimization, which takes a long time to solve and cannot guarantee global optimality.

There are also some methods that use vertical vanishing point information for video distortion technology. However, these methods do not consider the continuity between adjacent frames of the video and the jitter of the video, and the visual effect of the processed video is not good.

Images