RTI is a computational photographic method that collects between 40 and 100 images with light sources at different fixed locations. Once the software is able to deduce the location of the light sources, the resulting images can be used to create a mathematical 3D map of the surface.
Software can calculate surface normal vectors from the images. Subsequently, surface features can then be exaggerated to show the exquisite detail. The RTI technique is dependent on taking a number of images with light sources at fixed locations.
The process described in this article outlines a simple device that automatically collects reference images. The software has been developed by a number of researchers and is hosted by Cultural Heritage Imaging. Photographers, researches, and other interested parties are encouraged to read up on the technique. The construction and viewing of the RTI images is done with the RTI builder software maintained by the Cultural Heritage Institute and the resulting RTI file is viewed with their RTI viewer software.
The requirement of the light sources for the RTI image set is that they be about 50 in number and placed evenly on a half sphere.
One readily available dome was a simple lighting dome. In order to remove unwanted room light, I spray painted the inside black and then drilled 4 rows of holes evenly spaced from the horizontal. Each row holds 12 individual lights, which brings the total lights to 50 in number. Each hole was numbered to make the wiring easier. The large circular hole at the top of the dome is for the camera and was cut with a Dremel Drill. This technique easily removes the unwanted plastic and makes a smooth cut.
The lights are large 10mm white LEDs, which draw 80mA. These can easily be powered by the 5V DC supplied from the microprocessor. To control each LED, I used an LED light array manufactured by Adafruit Industries, which has the advantage of limiting the number of control wires needed to drive the 50 individual LEDs.
The wiring schematic is simple, but the execution is quite time consuming. It took about 12 hours to cut and solder all the contacts — this was slower than usual due to the Teflon coated wire, which was difficult to strip. Future systems will be assembled with a wire wrap tool, which is preferred for prototyping circuits.
The idea for the program is relatively simple too. Initially, an individual light is turned on, the camera shutter is triggered, and then the light turns off. The program sequentially repeats the cycle for all the remaining lights. The duration of the LED light is adjusted in the software. The prototype is designed for a 105mm macro lens and a Canon 5D Mark III DSLR camera body. In order to image at a low ISO and an adequate depth of field, the lights are programed to stay on for 5 seconds. This allows the full set of images to be collected in about 4 minutes.
The exposure can be optimized for the subject and increased if desired. If a large number of RTI image sets are being collected, the LED on time should be reduced to just slightly longer than the exposure time. To ensure repeatability and exposure efficiency, it is important that the LED illumination continues for the complete duration of the shutter exposure.