Integrating Stereo with Shape-from-Shading derived Orientation Information
I've finally had a paper published, so thought I better put it up. I had to do a poster session, so I've also put up the poster; this is mainly because unlike the actual paper I had space for a diagram showing data flow, which makes things marginally easier to understand.

As a quick summary the paper is dealing with the combining of stereopsis and shape from shading, using Gaussian belief propagation for the core component. Its rather esoteric, and the approach as a whole is not exactly simple. But the core algorithm and ideas are not that hard really - I don't think a simpler working approach to this problem is that likely. Of course, the usefulness is rather limited as its constrained by the limitations of both approaches, but it is (I hope) a small but valid step to something actually useful.

Integrating Stereo with Shape-from-Shading derived Orientation Information.

Poster of above.

It may be found in the British Machine Vision Conference 2007, Vol. 2, pp 910-919. The BMVC website will also have a copy, in case there is a problem with my website.