I believe making the algorithms Open Source is the best way to share research results in computer science. It happens too often that clever algorithms are rarely used and known because nobody is willing to spend weeks implementing the method from the paper.
To make new research even more accessible, I try to organize the code into open source projects, which let the others contribute and build their work on top of the existing code. This has been quite successful approach so far: I got plenty of feedback from the user community (bug fixes, performance improvements, better algorithms); pfstools have been included into several major Linux distributions; and I am amazed by the number of people that have fun when playing with the software.
These are the main projects that I started, contributed to or maintain:
pfstools package is a set of command line programs for reading, writing and manipulating high-dynamic range (HDR) images and video frames. It includes Qt and OpenGL HDR image viewers, tone-mapping operators and the tools HDR merging (camera response curve recovery). pfstools can be integrated with GNU Octave or matlab, so that it can serve as a toolbox for reading and writing HDR images.
High dynamic range visual difference predictor (HDR-VDP) is a near-threshold visual metric for comparing both scene-referred (HDR) and output-referred (LDR) images. The outcome of the metric is a probability of detection map, which tells how likely people will notice difference between two images.
Luma HDRv is an Open Source library and a set of tools for compression of HDR video. The codec has been designed to efficiently encode all visible colors and high dynamic range while ensuring that the visibility of errors due to coding is minimized. The stream is compressed using Google's VP9 video codec, which provides a 12 bit encoding depth.
pwcmp a set of matlab functions for Bayesian scaling of pairwise comparison experiment results under Thurstone's model V. The software can work with imbalanced or incomplete experiment designs; computes confidence intervals; improves accuracy for small sample sizes; can perform outlier analysis.