Presentation 2006/9/20
Image Rendering under Any Light Source Direction by Learning of Surface Reflectance with Neural Network
Haruki KAWANAKA, Yuji IWAHORI, Robert J. WOODHAM, Kenji FUNAHASHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) We have proposed Neural Network Based Rendering to generate a realistic virtual image of an object under any light source direction and any viewing direction, according to the physical reflectance property of an actual object. This method can generate a realistic virtual image of an object which has non-Lambertian reflectance without assuming some parametric reflectance function. However, the method requires many learning data by taking the various phase angles to achieve the effective learning of the neural network, that is, many images are required to learn the reflectance property with changing the light source directions. In the previous method, we used the angle between the light source direction and the viewing direction as one of the input data. Instead, the proposed method uses the angle between a unit vector which has the specular reflection to the viewing direction for each light source direction and a surface normal vector at each point of the object. The number of the images (light sources) can be reduced and the dense sampling for the learning can be achieved.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Neural Network / Virtual Image Generation / Mixed Reality
Paper # SIP2006-73,SIS2006-36,SP2006-59
Date of Issue

Conference Information
Committee SIS
Conference Date 2006/9/20(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Smart Info-Media Systems (SIS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Image Rendering under Any Light Source Direction by Learning of Surface Reflectance with Neural Network
Sub Title (in English)
Keyword(1) Neural Network
Keyword(2) Virtual Image Generation
Keyword(3) Mixed Reality
1st Author's Name Haruki KAWANAKA
1st Author's Affiliation Faculty of Info. Sci. and Tech., Aichi Prefectural Univ.()
2nd Author's Name Yuji IWAHORI
2nd Author's Affiliation Faculty of Eng., Chubu Univ.
3rd Author's Name Robert J. WOODHAM
3rd Author's Affiliation Dept. of Comp. Sci., Univ. of British Columbia
4th Author's Name Kenji FUNAHASHI
4th Author's Affiliation Grad. School of Eng., Nagoya Institute of Technology
Date 2006/9/20
Paper # SIP2006-73,SIS2006-36,SP2006-59
Volume (vol) vol.106
Number (no) 262
Page pp.pp.-
#Pages 6
Date of Issue