Yoldas

YOLDAS
- A raytracing rendering engine -

Thin Lens

Thin lens camera is the alternative camera of perspective camera. It has the adjustable aperture radius and focal distance, which enables depth of field effects and enhances the realism of our scene. There are more about realistic cameras here in the PBRT book.
The aperture radius is the size of the lens which controls how much light to enter. A larger radius leads to a shallower depth of field and more blur in out-of-focus areas. And the focal distance descibes at which distance the image is sharp.

Picture from left top to bottom right:
• with perspective camera
• thin lens camera with radius 0.05, focal distance 3
• thin lens camera with higher focal distance (7)
• thin lens camera with higher radius (0.09)

Halton Sampler

Halton Sampler is a better random number generator to reduce noise. It generates low-discrepancy sample points that are more evenly distributed than those produced by an independent sampler , which generates number with a simple pseudo-random generator.
We decided to do this feature because it looks pretty useful. The PBRT book helped a lots.

The left images are with the independent sampler, and the right are with Halton sampler with 128 sample points. You can see the difference in details with the lower images.

Denoising

Intel Open Image Denoise is an open source library of denoising filters for images rendered with ray tracing. It needs a normal image and an albedo image, and the output of the ray tracer to create the wanted denoising result. You can see the images with denoising feature, the quality is highly improved.

Rough Dielectric

Rough dielectric is another type of BSDF. With this feature, we can simulate frosted glass and other type of dielectrics by composing different value of roughness. Other factors such as ior, reflectance, and transmittance will decide the material of the surface.
The images on the right are the bunny with different reflectance and transmittance values.

Shading Normals

There are some textures contain extremely fine-grained details, which will become very expensive for our renderer if modeled directly. By applying normal maps, we can simulate these small surface features without increasing the amount of polygon too much.

Alpha Masking

Alpha masking simulates complex structures more efficiently without overloading geometry. It is ideal for objects with small and details features. Instead of using a detailed mesh to represent objects like leaves, alpha masks selectively remove parts of the meshes. If the alpha value is 0 (transparent), the ray passes through.

Third Party Assets