The Adaptive Path Tracing Framework works as a tool for developing and testing adaptive path tracing strategies via MATLAB®.
The project runs all of its computations on a pre-computed image sample data file. This helps to avoid performing path tracing steps by the renderer and makes adaptive path tracing simulation independent from specific renderer implementations.
The source code is structurised in scripts to avoid passing large data structures between function invocations.
The core script render.m takes input parameters from parameters.m, then reads sample data file and performs specified adaptive sampling and image reconstruction algorithms.
The project uses functionality of MATLAB® Image Processing Toolbox. It also takes OpenEXR as an external library to enable writing images in HDR compatible format.
The sample data file contains a list of samples. Each sample is represented by a tuple of values in binary
(pixel_X, pixel_Y, offset_X, offset_Y, colour_X, colour_Y, colour_Z, alpha)
Renderer "Mitsuba 2" was altered to enable printing out the sample values. Please refer to project "Path Sample Extraction" for generating the sample data files.
The code is invoked by calling the render.m script from project's root directory:
run('core/render.m')
All specific parameters have to be passed via parameters.m source file.
The project is available under MIT licence.