Finest Quality Superior Workmanship

This image has an interesting back-story and I imagine that a few of you probably caught the reference in the name. If you didn't; never fear, it just means that you probably haven't watched the movie Blade Runner as many times as some of us. It's a bit of a touchstone for both sci-fi films and the design community, and it's not a bad flick to watch either.

For those not familiar with the film, Deckard, a bounty hunter/cop and protagonist of the story, is trying to determine the origin of a scale he has found during his investigation. He takes the sample to a vender in synthetically replicated animals who analyzes it with a scanning electron microscope. Not only does she determine the type of animal it came from but also the manufacture, as the sample is tagged with a maker's id number. While not looking very snake skin like the image and idea behind the sequence was memorable.

I had been interested in scientific imaging for several years when I started playing with replicating the look of scanning electron microscope (SEM) imagery in CG. As a result, I had been collecting interesting images off the net as well as looking for books and other publications that had something to offer along theses lines. There are a few books currently available about SEM imagery. One of the best from the laymen's point of view/reference image collector is Journeys in Microspace : The Art of the Scanning Electron Microscope - Dee Breger

When the idea to replicate the Blade Runner image popped into my head I wanted to track down the original image if possible. The image in the film was naturally retouched to add the ID band and then only seen on a fairly low quality monitor on set.

The screen cap was serviceable but I figured the original had to be out there somewhere. About this same time I had stumbled across a long out of print book by David Scharf called Magnifications: Photography With The Scanning Electron Microscope.

To my great surprise on page 97 was the exact image I was hoping to find. Ironically the image isn't from an animal at all but a Cannabis flower.

While recreating the SEM shader effect was fairly straightforward the geometry was to be a bit more challenging. I figured that the general form was best done as a low ploy mesh to be smoothed out at render time with Meshsmooth.


Here you can see the base mesh with no additional levels of Meshsmooth applied.

The trick was how to get the very pro conspicuous nodule covered surface. My first thought was to grab individual faces from the low poly mesh that had one level of tessellation applied and extrude them slightly. When the render time Meshsmooth was added this would turn the box-like extrusions into rounded bumps. By varying the amount of extrusion and taper some degree of irregularity could be given to the surface while at the same time allowing a great deal of control as to where the bumps would be located. Unfortunately, after running a test on a section of the geometry using this method it became apparent that the uniformity of the underlying mesh would not easily be disguised. Additionally the original image had a good deal of variety in the size and placement of the nodules that would have necessitated rebuilding the base mesh to accomplish the same feel in the finished image. My second take on the surface was much less labor intensive and more straight forward. I had considered using the cellular map for displacement of the surface original but had dropped it for a couple of reasons. Firstly, while the cellular map would produce nearly a dead ringer for the majority of the surface I wasn't sure how I would be the more stretched out bumps visible on the stalks without masking them and blending them into the base surface. Secondly, Max really falls down dealing with displacement and I wasn't sure if I would be able to tessellate the surface to the desired resolution without running into memory limitations. As it turned out both of these problems were solved fairly easily.

Since the base geometry had been built from a geosphere I figured that instead of basing the displacement off of the XYZ of the displacement map, I could use UVW space and spherically map the surface. This then allowed me to set the tiling of the map's W component to something less than 1.0, thereby stretching the map along the normal of the original sphere. Doing this allowed the stalks to have longer, stretched nodules along their length without having to figure out a way to blend two different displacements between the base and the stalks. As for the displacement it was all accomplished with the Displace modifier and a high level of Meshsmooth. The Displacement modifier as it turns out is much quicker and versatile to use than the displace map within the material. Until the day when micropolygon displacement is built into Brazil this is a rough but passable solution.

The stalk caps were created from spheres squashed, manipulated and displaced with hand painted displacement maps and the filaments in the distance are simple splines turned into geometry with Clay Studio Pro.

Lighting and shading were straightforward as I had already come up for a fairly good recipe for the SEM look on previous images. The basics of the look are produced through a falloff map where the edges are illuminated and normals closer to the camera view fade to black. The exact appearance of the falloff varies from photo to photo and can easily be changed using the mix curve controls. Subtle bumps can be added for additional detail. Lighting can almost be non existent if self illuminated materials are used but for a closer match to the real thing a strong light and simulated bounce is necessary. Brazil's skylight was just what was needed.

Rendering was accomplished with an early version of Brazil. I wasn't sure how I was going to do the Depth of field based blur, but fortunately right about that time DOF was added in and allowed the background to be nicely out of focus just like the reference image. Just a hint of film grain was added in post and it was done.