The primary results from this project are the two Matlab scripts, hdrEnc and hdrSimulate, which allow for the encoding and simulation of images for the BrightSide HDR display. hdrEnc matches the output of the BrightSide encoder script closely, and runs in a third of the time. It optionally allows for a convex optimizer solver to be used to reduce MSE between the encoded result and the source HDR image, although with the cost of drastically increased runtime.
The convex optimizer reduces mean square error by a factor of 2 to 10, depending on how well the source image fits into the dynamic range of the display. The increased runtime, however, makes it infeasible to use for large amounts of data, and it is certainly not suitable for real-time applications. Finding such an encoder would also be interesting, as the original encoder also runs far slower than needed for video output.
hdrSimulator is used to validate, quantify, and support the encoder, and is in general a useful tool for exploring the capabilities of the HDR display, and for further study into HDR encoding. As more detailed parameters for the display are measured, they can be plugged into the simulator to further improve its fidelity. Note that the convex optimizer ties into the simulator in a deep way, since the simulator is used to generate the mapping matrix for the optimizer. Therefore, the output from the CO encoder will improve as the simulator fidelity improves.