aromanro / raytracer Goto Github PK

A ray tracing program

License: GNU General Public License v3.0

C++ 100.00%

ray-tracing raytracer raytracing raytracing-one-weekend ray-tracer wxwidgets wxwidgets-applications optics raytracing-render 3d

raytracer's Introduction

RayTracer

A ray tracing program

Mainly based on 'Ray Tracing' books from Peter Shirley: http://in1weekend.blogspot.com/2016/01/ray-tracing-in-one-weekend.html I skipped the boring (for me) texture generation part and also the motion bluring, I preffered to go further enhancing the materials and loading obj files.

Some more info is on Computational Physics Blog: https://compphys.go.ro/ray-tracing/

Some papers that were used:

'An Anisotropic Phong BRDF Model' by Michael Ashikhmin and Peter Shirley https://www.cs.utah.edu/~shirley/papers/jgtbrdf.pdf

Osada et All, 'Shape Distributions' http://www.cs.princeton.edu/~funk/tog02.pdf

'Fast, minimum storage ray-triangle intersection' Tomas Möller and Ben Trumbore. Journal of Graphics Tools, 2(1):21--28, 1997. http://www.graphics.cornell.edu/pubs/1997/MT97.pdf

The project has some known issues, for example loading an object from an obj file does not split correctly concave polygons, they must be convex (unless you are lucky). See the code for how that could be fixed. Also importance sampling with composite objects has some issues. I know why they occur and how it could be solved, but I don't have enough patience for those changes right now :)

Obj loader is very basic, it seems to work for some cases.

An animation generated with the program:

Some images obtained while implementing it:

First, with more recent code, an obj that has specular mapping and a normal/bump map:

After the first book, 'Ray Tracing in one weekend':

With local illumination, depth of field and textures:

The Cornell box (during the 'next week' book):

Adding triangles, already beyond the 'rest of your life' book:

Adding Beer-Lambert law for transparent objects:

The first obj file loaded in the Cornell box:

Making it transparent:

Trying some other obj models:

Now together with colors/textures:

Implementing the Phong model:

Trying it on a model, exponents are too high:

More testing:

And some new model for testing the features:

Images generated while fixing some bugs in models loaded from obj files:

And one after some later changes, when I also added procedural textures:

raytracer's People

Contributors

Stargazers

Watchers

Forkers

lp249839965 phantom7knight shadercoder fieldplay mmmagdalena zmy12006 leeqmoon nguyenvuducthuy mbrukman jerryzfc generalzzd tanganke horew lelouchd frederickmo berysaidi wahello

raytracer's Issues

how to improve it？

Add some procedural textures

I kind of skipped that part of the book, I considered it boring.

Improve splitting polygons into triangles

I think that could be made to work better, now it does it wrong for some models, in some parts.

Improve 'materials'

There is a lot to be improved here.

A good book on the subject is available for free online: https://pbr-book.org/

Improve importance sampling

Currently importance sampling works only for simple objects (that is, not composite). For this project it's enough, it's just a proof of the concept and only simple ones are needed (as lights, mirrors), but it could be enhanced to work for more complex stuff as well.

Anisotropic Phong

Hi, I've been working through the Ray Tracing in One Weekend book. When trying to implement a Anisotropic Phong BRDF I had bugs in my code which I couldn't immediately fix until I saw your project and how you implemented it. I used a few lines from your project to get mine to work. As your project uses the GPL licence, I was wondering if I could have your permission to leave the lines of code in the Anisotropic Phong BRDF, where it looks similar to yours, in my project and release my project under the MIT licence? I've given you credit by linking to your github Ray Tracer project both in the ReadMe file and in the code itself. If you don't mind me releasing under the MIT licence, I'm happy to credit you in any way you like for example by also including a link to your website and your name, whatever you would be happy with. I haven't released it yet until I sort out the licence.

Some parts of some models are wrong

Parts of some models that are composed of triangles come out wrong (so it's not due of: #7).
I don't know yet why, might be combined with the texture mapping being wrong.

Add skysphere

Similarly with what is already implemented in the https://github.com/aromanro/SolarSystem project.

Obj file loader: normals can be missing

In such a case, they could be computed for each vertex using the triangle vectors.

Avoid loading the same texture twice

Many obj files have set the same texture file for different parts of the object and/or the same one for both ambient and diffuse textures. Avoid loading them multiple times by using a 'cache', first check in there and if it's not already loaded, then load it.