Single-image understanding under unknown radiometric response functions

2016 Fiscal Year Final Research Report

• Project/Area Number: 26540088
• Research Category: Grant-in-Aid for Challenging Exploratory Research
• Allocation Type: Multi-year Fund
• Research Field: Perceptual information processing
• Research Institution: Kyushu Institute of Technology
• Principal Investigator: Okabe Takahiro 九州工業大学, 大学院情報工学研究院, 教授 (00396904)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: コンピュータビジョン / イメージベーストモデリング

Outline of Final Research Achievements

In this study, we have developed an image-based technique for recovering the description of a scene such as shape, reflectance, and illumination from a single image taken by a consumer camera with an unknown and nonlinear radiometric response function. Specifically, we proposed a joint estimation of an illumination distribution and a response function on the basis of their low-dimensional linear models, a joint estimation of surface reflectance and a response function via entropy minimization, and a joint estimation of an illumination distribution, surface reflectance, and a response function by combining the above two methods. In addition, we investigated how the surface shape recovered by conventional shape from shading is distorted by nonlinear response functions, and tried to jointly estimate surface shape and a response function on the basis of the integrability constraint.

Free Research Field

視覚情報処理