Semi-Automatic 2D to 3D Video Conversion
| Autor: | Ekaterina V. Tolstaya, Petr Pohl |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Matching (graph theory)
Computer science business.industry Hash function ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Inpainting Optical flow 2D to 3D conversion Motion vector Feature (computer vision) Key frame Computer vision Artificial intelligence business ComputingMethodologies_COMPUTERGRAPHICS |
| Zdroj: | Smart Algorithms for Multimedia and Imaging ISBN: 9783030667405 |
| DOI: | 10.1007/978-3-030-66741-2_4 |
| Popis: | In this chapter, the process of converting conventional video sequences to stereo format is considered. Key frame detection, depth assignment, depth propagation, motion vector estimation, background inpainting, and stereo synthesis comprise the stereo conversion pipeline. We also consider several aspects of content production, such as stereo shooting and stereo conversion. The depth propagation algorithm is based on patch matching using patch hash codes, and the results are compared with motion vector-based depth propagation results. Motion vector estimation is an essential algorithm that is used to reveal video clip attributes, and it plays an important role during conversion. We propose an optical flow algorithm based on a primal-dual optimization algorithm. This work is based on the authors’ experience during a large-scale project at the Samsung Research centre in support of newly appearing TVs on the market, which were equipped with a “3D-ready” feature. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|