Aggregation of Rich Depth-Aware Features in a Modified Stacked Generalization Model for Single Image Depth Estimation

• Published in: IEEE transactions on circuits and systems for video technology 2019-03, Vol.29 (3), p.683-697
• Main Authors: Mohaghegh, Hoda, Karimi, Nader, Soroushmehr, S. M. Reza, Samavi, Shadrokh, Najarian, Kayvan
• Format: Article
• Language: English
• Subjects: 2D to 3D image conversion; Computer vision; data-driven approaches; depth estimation; Estimation; Feature extraction; modified stacked generalization model; monocular cues; Qualitative analysis; Regression analysis; Semantics; Solid modeling; Three dimensional models; Three-dimensional displays; Training; Two dimensional displays; Two dimensional models
• ISSN: 1051-8215; 1558-2205
• DOI: 10.1109/TCSVT.2018.2808682

Estimating scene depth from a single monocular image is a crucial component in computer vision tasks, enabling many further applications such as robot vision, 3-D modeling, and above all, 2-D to 3-D image/video conversion. Since there are an infinite number of possible world scenes, that can produce a unique image, single image depth estimation is a highly challenging task. This paper tackles such an ambiguous problem by using the merits of both global and local information (structures) of a scene. To this end, we formulate single image depth estimation as a regression problem via (on) rich depth related features which describe effective monocular cues. Exploiting the relationship between these image features and depth values is adopted via a learning model which is inspired by modified stacked generalization scheme. The experiments demonstrate competitive results compared with existing data-driven approaches in both quantitative and qualitative analysis with a remarkably simpler approach than previous works.