In the end, I think the advantage of the deformation map approach when compared to free-hand sculpting gives very accurate control of the placement of details, especially when layering several deformation maps. The details fit the colour texture nicely. The downsides are the need to paint many deformation maps (about ten for the body alone), and having to switch back and forth between ZBrush and the paint program for every change in deformation. Worst of all, there was visible UV-stretching when deformations became too large.
After detailing, I went back into Maya and set up an underwater photograph back-plate, and matched the lighting to it. The back-plate can be seen in Fig.15.
The Volume light setup used Mental Ray physical light and the Mental Ray participating Volume shader, applied to a cube. The spheres in front of the Volume light were partially transparent, because of the visible streaks in the light cone. There was also a Fill light from below, and an IBL-gradient in the scene. The white sphere next to the head was not renderable and was used for the eye reflection only. The Volume light setup in Maya view port can be seen in Fig.16, and the rendering result in Fig.17.
I re-imported the "cage object" models from ZBrush and smoothed them in Maya, by one level. I used the Mental Ray architectural material for all shaders, as non-Mental Ray shaders are known to not work well with the physical light. I applied the painted colour texture and the normal map. The cavity maps went into the extra colour as a multiplier. The re-imported and smoothed 1 or 2 level cage objects can be seen in Fig.18.
I rendered everything in Mental Ray in 2K resolution and 16-bit colour depth. Because it's a still image, there were several passes for the lighting. I then put them all together in Photoshop. Some render passes: projected caustics and self shadow passes can be seen in Fig.19 - 20; Volume light and Ambient Occlusion passes in Fig.21 - 22; and Specular and Reflection passes in Fig.23 - 24.
I had to do some painting for the generated normal maps and on UV seams in the finished image, where I had not ideally placed the position of the camera. I finally applied depth of field using a Z-map, and used a Photoshop lens blur. The result can be seen in Fig.25.