Hacker News new | ask | show | jobs
by jb1991 1041 days ago
Since modern 3D renderers are already designed to render in a photorealistic way using fundamental principles of light transport, what does this technique actually offer that the renderer itself does not offer? Path tracing follows the basic laws of global illumination, after all, and the path tracing algorithm offers simulation of all different kinds of lens types, etc.
3 comments

Partly this is a fun exploration of creating art in an unusual medium, much like a photographer might experiment with antique lenses, unusual film formats and suchlike.

But also, although a lot of modern 3D renders are highly realistic there are certain levels of realism they don't bother to pursue, for both computational efficiency reasons and the fact a lot of users don't want them.

This chap had to simulate a lot of light rays, because only a small fraction of them ended up on the imaging plane. You can achieve effects like bokeh adequately with much less render time by using a conventional blender camera and a blur proportional to the Z buffer.

Meanwhile things like film grain, the corners of the image being darker, spherical aberration, chromatic aberration, parts of the image out of focus, lens flares etc aren't what everyone wants for their image.

My understanding is that modern renderers render in a "physically accurate" way. However, cameras throughout history has never captured images in a "physically accurate" way but rather through a set of imperfect techniques (lenses, film etc) which produces a certain look. When rendering, it's hard to make things seem realistic as they often look too perfect but something like this gets closer to how a photo would look.
Path tracing isn’t really physically accurate, as it also models how cameras work. Path tracing is considered “photorealistic “. I’ll have to dig into this project more, but all pathtracing algorithms start with the fundamental principles of optics in the lens, and changing the field of view of that lens affects the image and the perspective distortion, and you also have effects like the depth of field, and other aspects we associate with cameras. Depth of field, after all, is not a physically accurate phenomenon but rather based on how cameras work, and yet that is also how 3D renderers interpret the environment.
But those algorithms generally implement only the most basic aspects of a lens.

No one normally implements a model of a cheap lens that has severe chromatic aberration due to poor focusing of different wavelengths in its poorly designed glass stack.

Depth of field also exists in eyeballs too
It’s not quite the same effect as a camera. Eyes also do things that cameras do not.
This is an art project. A breathtakingly technical art project.

Did you expect this to be practical? It's a fantastic hack.