1. Fast elastic motion estimation with improved Levenberg-Marquardt optimization.
- Author
-
Song, Chuan-Ming, Min, Xin, Sun, Shiqi, Wang, Xiang-Hai, and Yin, Bao-Cai
- Subjects
- *
HESSIAN matrices , *VECTOR spaces , *VIDEO coding , *ADAPTIVE computing systems , *ERROR functions , *GAUSS-Newton method - Abstract
• This study presents a practical optimization framework of elastic motion estimation. It employs the Levenberg-Marquardt method to facilitate fast gradient descent speed at early stage of iteration by introducing a diagonal matrix into the Gauss-Newton's Hessian matrix. Thus, the slow convergence of conventional Gauss-Newton's optimization method can be efficiently dealt with, especially when the initial solution is far from the optimum. • We prove that the Gauss-Newton's Hessian matrix is positive semi-definite with a probability of 0.6951 on average, by discussing the convexity of the motion-compensated error function. Such an observation lays a theoretical foundation for us that we can adopt a sign-alternating factor, instead of a positive constant used in the traditional L-M method, to update iteratively the diagonal matrix coefficient. This strategy guarantees that our method searches for the optimal motion vector in a much bigger parameter space. • To avoid a negative diagonal matrix coefficient producing a singular or ill-conditioned Hessian matrix, we further determine an upper bound and a lower bound for the diagonal matrix coefficient, with whose constraints we propose an adaptive method to compute the sign-alternating factor to restrict the update process into a reasonable range. • The proposed elastic motion estimation algorithm explores simultaneously a bigger parameter space and a more suitable update manner which effectively adapts to the distribution of the motion-compensated error function. Therefore, it outperforms the full search based on block-wise translational model and the elastic motion estimation based on Gauss-Newton's method in terms of the motion-compensated peak signal-to-noise ratio (PSNR). Furthermore, our method needs only 1–2 iterations before it achieves higher PSNR than conventional elastic motion estimation and the block-wise translational full search. The elastic motion estimation is an efficient temporal predictive coding tool for predicting complex motions including zooming, shearing and twisting. However, its optimization solution still suffers from high computational complexity and unstable convergence yet. We thus present a novel optimization framework to improve the elastic motion estimation efficiency. By introducing a diagonal matrix into the Gauss-Newton's Hessian matrix, we employ the Levenberg-Marquardt optimization to compute the elastic motion vectors. It obtains superior convergence to the conventional Gauss-Newton's method, especially when the starting point is far away from the globally/locally optimum. Furthermore, we prove that the Gauss-Newton's Hessian matrix for elastic motion estimation is positive semi-definite by discussing the convexity of the motion-compensated error function. Such a finding allows us to adopt a sign-alternating factor, instead of a positive constant used in the traditional L-M method, to update iteratively the diagonal matrix coefficient. And a bound is determined for the diagonal matrix coefficient to guarantee the non-singularity of the Hessian matrix, with whose constraints we propose an adaptive method to compute the sign-alternating factor to restrict the iterative process into a reasonable motion vector space. Finally, a novel elastic motion estimation algorithm is presented based on the modified Levenberg-Marquardt optimization. Experimental results show that the proposed algorithm gains higher motion-compensated peak signal-to-noise ratio (PSNR) than several benchmark methods. Moreover, our algorithm converges fast which needs only 1–2 iterations before it achieves higher PSNR than conventional elastic motion estimation. [ABSTRACT FROM AUTHOR]
- Published
- 2022
- Full Text
- View/download PDF