Abstract:
Reflectance Transformation Imaging (RTI) is a technique that retrieves an object's per-pixel reflectance behavior, capturing different images of the object illuminated by different known, or knowable, light directions.
The purpose of RTI is to analyse the surface of an object and from that tries to reveal details about the object that are difficult or impossible to reveal from manual and human investigation.
To achieve this RTI has been proposed with different techniques to interactively relight the object, simplifying the way to study an object surface.
Initially RTI was difficult to implement due to the required equipment to carry out it and the different setups involved to retrieve the different light positions used in the process, making RTI only suitable inside a lab.
Nowadays, however, in the literature, many other alternatives have been proposed for the acquisition and processing of the data to carry out RTI, involving new tools and methods.
For this reason, we propose a novel approach in which RTI is carried out using only two smartphones recording a video of the object, where the first one uses the flashlight mounted near the camera to illuminate the object from different position, while the other one is placed in front of the object to monitor its reflectance.
Then, to deal with the data from the two videos we propose a neural network responsible to reconstruct the object reflectance for any light source, which model will be used as part of the interactive relighting of the object.
Lastly, we propose a simple application where any user can try the RTI method we propose, providing a simple UX to create the interactive relighting model for any experiment video made with two smartphones.
