Reflectance Estimation and Dataset construction using multiple light sources

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K06076
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Communication/Network engineering
• Research Institution: The University of Kitakyushu
• Principal Investigator: Okuda Masahiro 北九州市立大学, 国際環境工学部, 教授 (10336943)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: 画像復元 / 画像分離 / 反射率推定 / ハイパースペクトル画像 / 高ダイナミックレンジ画像

Outline of Final Research Achievements

① We devised a method of separating an image into the reflectance and the shadow image by using flash and no-flash images. Both subjective evaluation and objective evaluation showed that the performance is superior to conventional methods. Also, by developing this approach, we proposed a method to reduce blur in the dark by combining the image using flash and the image taken with high ISO sensitivity. We formulated the image integration problem using Local Linear Model and succeeded in effectively solving by using convex optimization method.
② We proposed a method to obtain the reflectance image at high speed by using the spectral information of each pixel.
Three years of external presentation results were 9 journals of academic journals and 29 conference presentations.

Free Research Field

信号処理，画像工学