specular i


£ c

lighting Ano


hont ons Fo fOOIc hont ons Fo fOOIc and photograph a subject m a way that eliminates r single photos the best we can do is use uniform - - m 'S that a lot of the green and most of the red light ergy )S trom sub-surface scattering and not diffuse surface scattering we win just look at suppressing the specular component rujiunaieiy. skin has very little reflectance normal to the surface (looking onto fne sk,n>. but't is very reflective at low angles. It we take a whole ser or pictures under the same uniform lighting, but at different angles, we can do a good job of suppressing the low-angle specular component

For now Fortunat

O Step 7: Correcting the diffuse for side-lighting

At least we have a nice flat-lit blended map to start with. The contamination with specular light is fairly even, and we can easily remove it for a good approximation. We take a sample of the skin color in the photograph at a point that is free of strong specular component, but not so low an angle that we are starting to get falloff. Gaussian blur the images first and then

take a 5-pixel box average in the cheek area. We will also take hi point-sample in the pupil inside the reflection shadow of the photoo h and a mid-sample in the lower iris area. You correct the blended3 using these comparative samples. maP

To do this the shader weights each of the blended photographs by the <m, product of the surface normal and the camera angle We do this on a in? frequency (softened version) of the photographs and use this RGB VaZ' to correct the high-frequency (full resolution) photographs. Looking at m! ohotograph on the left, and the blended maps on the right we see that 7 nave managed to flatten the lighting Unfortunately, the photographs W1 taken with strong side-lighting-see the reflections in the eyes. This si „anting results in strong specular highlights towards the camera. So we up with a very uniform blended map with a significant specular compone t


Skin deep h

0 0

Post a comment