Unguided Depth Map Completion Using Weighted Schatten p-norm Minimization
| Autor: | Sukla Satapathy, Rajiv R. Sahay |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Matrix completion
Matrix norm 02 engineering and technology Inverse problem 01 natural sciences Regularization (mathematics) 010309 optics Depth map 0103 physical sciences Convex optimization 0202 electrical engineering electronic engineering information engineering 020201 artificial intelligence & image processing Minification Convex function Algorithm Mathematics |
| Zdroj: | ICCE |
| DOI: | 10.1109/icce50685.2021.9427715 |
| Popis: | As depth map completion is an ill-posed inverse problem, the regularization constraint plays an important role. Matrix completion with nuclear norm minimization and a data fidelity term with Frobenius norm leads to a convex optimization problem. However, its solution is approximate and different from that of original rank minimization problem. In this work a weighted regularization method for depth map completion without using its associated RGB image is proposed which involves a non-convex rank minimization process utilizing the weighted Schatten p-norm(0 < p ≤ 1). A detailed experimental analysis shows that our approach, outperforms the state-of-the-art depth map completion approaches. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|