A Novel Approach for Reflectance Capture

Ladefoged, Kasper Skou

4. term

Vision, Graphics and Interactive Systems, Master

2016

2016-09-09

67 pages

Capturing the reflectance of an reconstructed object is of use in many area of computer graphics. Currently the only way to do this is using big, one of a kind, static setup. This limits the accessibility of reconstructed objects. Creating a more novel approach with, off the shelf hardware, would increase the presence of real world object in augmented reality, virtual reality etc. This project strives to achieves this, by capturing multiple images at the same locations in order to isolate a known light source, and then reverse engineer the physical object reflectance. This creates some guidelines in the capturing process, in order to ensure all the variables are inside a given range, in order to limit errors. The precision of the results primarily depends on the precision of the calibration of the flash, and the reconstruction of the physical object. In order to limit the error, a process for calibrating a light source is presented. Furthermore the reconstruction of the physical object is done with the use of ContextCapture, as this creates very high quality reconstructions, while presenting the cameras estimated positions and rotations in relation to the reconstruction. The acceptance test shows that the error of a synthetic test, has a mean error of −0.0816. Given a range for the pixel values of 0 to 1. The shadows produced by unknown light are removed, but a series of bright spots are introduced in the real world test. This has to be investigated further, but seems very promising

Computer Vision ; Vision ; Computer Graphics ; Reflectance ; Reconstruction ; BRDF ; Bird Bath ; Graphics ; IBM ; Image Based Modelling

Download
View record in AAU Student Projects