Loading...
Please wait, while we are loading the content...
Similar Documents
Calcul variationnel des dérivées d’une image et application à l’estimation du flot optique et du flot de scène
| Content Provider | Semantic Scholar |
|---|---|
| Author | Mathlouthi, Yosra Mitiche, Amar Ayed, Ismail Ben |
| Copyright Year | 2016 |
| Abstract | In computer vision, image derivatives are generally computed by local averaging of finite image differences. This paper describes a variational method of regularized differentiation to estimate the partial derivatives of an image, with an application to two important motion analysis problems : optical flow and scene flow estimation. This variational method minimizes a functional composed of an anti-differentiation data conformity term, and a classic smoothness regularization term. The data term constrains a derivative to be a function which, when integrated, produces the image. Discretization of the corresponding Euler-Lagrange equations yields a large scale sparse system of linear equations, which can be efficiently solved by an iterative method such as Gauss-Seidel. We ran qualitative and quantitative experiments, using real and synthetic images, which show that optical flow and scene flow obtained with image derivatives computed by regularized differentiation are more accurate than those obtained using standard derivative definitions by finite difference averaging. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://www.etsmtl.ca/Unites-de-recherche/LIVIA/Recherche-et-innovation/Publications/Publications-2016/P06.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
