Rendering Mental Ray: Anti Aliasing Settings

Anti-aliasing quality

One important part of the quality of the rendered Mental Ray image is the anti-aliasing of the image. This will smooth out the artefacts, which will appear at the edges of surfaces and on the textures you use.

Jagged or stair-cased edges in pixel-based images or flickering surfaces in an animation are aliasing artifacts.

Aliasing artifacts result from point sampling, a process used in all computer graphics applications that determines the information about each pixel. Aliasing artifacts can result at various stages during any rendering process.

Anti-aliasing is the process of removing or reducing these artifacts. Because there are many kinds of aliasing, such as grainy surfaces, flickering, and jagged edges, there are as many approaches to controlling or fixing these problems. You can make adjustments to a number of settings to decrease or eliminate aliasing artifacts and flicker.

Most solutions to control aliasing are time consuming and increase render times. Try to find the solution that gives you the best balance between image quality and performance.

Number of Samples

Sampling mode

Selects the sampling mode used. Change to Legacy Sampling Mode and leave Legacy Sampling Mode to Adaptive Sampling which is recommended for most cases.

Min Sample Level

This is the guaranteed minimum number of samples per pixel used when processing an image. Based on contrast sensitive (adaptive) settings, mental ray for Maya will increase these samples as needed.

Max Sample Level

This is the absolute maximum number of samples per pixel used when processing an image.

Test renders: -2 0 or -1 1 Final renders: -1 1 or 0 2

Anti-aliasing Contrast

If the contrast between samples is higher than the value set here, Mental Ray will increase the number of samples to get better results.

Test renders: 0.1 0.1 0.1 Final renders: 0.05 0.05 0.05

Multi-pixel Filtering

This is processing performed on the results of the sampling to blend pixels into a coherent entity. Black and white = noisy. Filtering looks at neighboring info and unifies the two.

Box Filter (default)

The fastest way to get relatively good results.

Triangle

More processor intensive than box, but offers even better results.

Gaussian Filter

Produces the best results, but is the slowest to render.

Mitchell, Lanczos

Mitchell (clip) and Lanczos (clip) are alternatives to Gaussian that offers slight variations in contrast (tends to increase). Mitchell increases less than Lanczos. Because "plain" Lanczos and Mitchell may produce negative values the new filter types are "clipped" variants to ensure positive values. The filtered result samples are clipped to the min/max range of input samples. The final pixel in the image will therefore not contain any out-of-range values _produced by the filter_, as might be the case for regular Mitchell and Lanczos filters.

Filter Width and Height: Gives the amount of pixels that will be used in the filter calculation. A higher value for the width and height will translate into soft edges in the image. Use the standard value if no artefacts at edges in the image are visible.

Test renders: Triangle or Gauss Final renders: Mitchell, Lanczos or Gauss

Sample Options

Jitter

Reduces the calculation errors which can occur with the sampling of the edges.