MICCAI 2022 Daily – Monday

The same also applies to non-Lambertian reflectance. For all three problems, the curve looks like this very flat curve. To convert this problem into maths, Jingwei uses Bayesian modeling . He gives the matched one some disturbances and estimates this kind of curve, using this as a probability. The probability is high for well-illuminated, well- conditioned figures, and for the other three problems, the probability is low. “ Here is the entire diagram of this process – it’s a coarse-to-fine process, and the left to right image patch matching is on all levels, ” he explains. “ Each level can be used as an initial value for the next finer level. You may know the optical flow algorithm named Lucas-Kanade, and this is a faster version called Fast Lucas-Kanade. I didn’t create it, but the idea is simple. The patch on the left should be matched on the right, and we should 6 DAILY MICCAI Monday Oral Presentation