Choosing optimal correction in video stabilization

摘要

To correct for the motion a perspective transform of a pair of matrices can be applied to an input image to provide an output frame. The first matrix can represent a transform to be applied to the input frame and the second matrix can represent an identity matrix. Each matrix can contribute to the output frame according to a respective weighting factor. The weighting factors for the two matrices can be determined based on an estimate of the overscan.

主权项

1. A video stabilization method, comprising:
generating an output frame from an input frame according to perspective transform of a pair of matrices, a first matrix representing a transform to be applied to the input frame and the second matrix representing an identity matrix, each matrix contributing to the output frame according to a respective weighting factor; and determining weighting factors for the two matrices based on an estimate of overscan.