11
$\begingroup$

The iris (the colourful ring surrounding the pupil of the eye) is covered in a layer of water, and appears to have opaque elements embedded in transparent and translucent elements. What effects do I need to model in order for the iris to appear realistic close up (the iris taking up over 20% of the image area)?

Is sub surface scattering necessary or is transparency sufficient? Is it necessary to take into account light from within the eye that entered via the pupil, or can the back of the iris be regarded as completely opaque? Are there other effects that I haven't considered?

I'm looking to produce still images offline - the approach does not need to work in real time.

So far I've tried covering the eye in a thin transparent film to simulate a water layer, and treating the iris as a section of a transparent ball with coloured transparent radially arranged strands, offset as they are layered front to back. To prevent unwanted effects from light that gets all the way through, I have backed the iris with a matt black sphere section. This still seems to give an eye that looks somewhat artificial (inorganic) and disconnected though.

I'm trying to make this work with geometric primitives such as spheres, cones and cylinders, but I'm open to an approach using triangle meshes if that opens up new possibilities.

$\endgroup$
1
  • 3
    $\begingroup$ Subsurface scattering is for sure going to be important in such a close-up. You might look at Jorge Jimenez's work for some inspiration; see his talks at SIGGRAPH 2012 and GDC 2013. His work is for real-time but I'm sure some of the ideas can be adapted. $\endgroup$ Aug 13 '15 at 6:13
12
$\begingroup$

For a realistic Iris, you need:

  1. Geometry: Irides are very different for different people and their appearance changes drastically depending on whether the pupil is dilated (sphicter muscle relaxed) or not. You will not get anywhere with primitives only. The whole structure is like a thick stretchy wrinkly cloth suspended in water. Wrinkles appear and disappear as it moves. Use a sculpting software and an anatomy textbook.

  2. Material: There is almost no specular on the iris since it is suspended in (basically) water which has a similar refractive index as tissue. I do not know the roughness values since I have never seen a dry Iris. With the GGX specular model, roughness 0.4 looks okay. Brown irides seem to be more rough, almost like very fine sand. A shading model that uses roughness in the diffuse component would be nice.

  3. Textures: While some are colorful, the vast majority of irides in the world are brown. You don't need too much variation inside the texture, it can be quite uniform. Make it relatively saturated and don't try to hint at geometry inside the texture by drawing dark areas for shadows or occlusion. Have a normal map. If the eye is animated, you need to blend between at least two normal maps for the dilated and undilated pupil. I don't think you need a roughness map, but especially for blue irides it might help.

  4. Refraction and reflection at the cornea: The cornea (the fluid layer coating it) acts as a lens. This distorts the image of the iris as seen by the camera. Without this, it will not look real, especially when looking at the eye from the side. The unusual thing about the eye is that, from a graphics point of view, the object is inside the lens. So there's no thin lens equation going to come to your rescue. You have to refract light once, at the interface between the tear layer and air. The other interfaces (tear layer to outside of cornea, layers and membranes inside the cornea, inside of cornea to chamber fluid) are not relevant since the indices of refraction are all very similar or the structures are extremely thin. The only other optically important structure is the lens, which is behind the iris and can be ignored. The shape of the cornea is critical, as it will distort everything. People with cornea deformations such as a keratoconus have weird-looking eyes. Also the whole scene must reflect on the cornea/liquid. We expect to see reflections of light sources over the iris/pupil.

  5. The limbus: Looking frontally at the eye, there is a greyish ring around the Iris. This is the part where the cornea merges into the sclera (the "white" of the eye), transitioning from transparent to nontransparent. Behind the limbus, anatomically, is the iridocorneal angle, where the iris meets the rest of the eye. This area is impossible to see due to the refraction mentioned above, it is literally a place that has never seen any light. Light emitted from there experiences total internal refraction at the interface of the cornea, you can't look at it from the outside. Thus, it looks very very weird when the transition from iris to the rest of the eye is visible. You need both the refraction and the gradual transition from cornea to sclera to get this transition right.

  6. Subsurface scattering: You might possibly get away without it on the iris itself since there will never be light coming from the back, especially with a brown iris (more heavily pigmented -> more attenuation during scattering). For some light directions, it will make a difference though. And you need it on the surrounding skin and sclera anyway. But it is not the most important effect.

  7. Ambient occlusion: You need good SSAO (or GI) to make the fine structure stand out. There's almost no specular, so you need other visual cues for the geometry. Choose a method that uses your normals from the normal map. HBAO will work well. For offline rendering, the GI will be done by your renderer of choice.

  8. Lighting: Using some physically based method goes without saying. Image based lighting, especially for the cornea reflection. Under certain angles, refraction of the light at the cornea may become relevant as well (only for analytic lights). Postprocessing (bloom), especially for the cornea reflex.

Backscattered light from the retina, although it can be seen through the iris in very extreme viewing conditions during some medical examinations, is completely irrelevant for normal purposes. Basically, paint the retina black. If you want to go that far, you can add a little orange-red to the retina color whenever a strong light comes from a direction very very similar to the camera and the pupil is dilated (like a flash from a camera). But it will still not be strong enough to shine through the iris, by far. Black is an okay approximation.

That list looks big, but you asked for photorealism ;) It will still look good if you only do the easy half and fake the rest.

(Source: I do graphics programming for medical simulations with a focus on the human eye)

Material:

$\endgroup$
2
  • 3
    $\begingroup$ Would be interesting to see some results of your detailed recipe if there are any you can share. $\endgroup$
    – Rotem
    Sep 22 '15 at 14:52
  • $\begingroup$ Not at the moment, sorry :/ $\endgroup$
    – cupe
    Sep 23 '15 at 13:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.