Antialiasing Filters

Introduction

In this chapter we'll cover the various anti-aliasing filters implemented in V-Ray. According to Wikipedia: "An anti-aliasing filter (AAF) is a filter used before a signal sampler to restrict the bandwidth of a signal to approximately or completely satisfy the sampling theorem over the band of interest. Since the theorem states that unambiguous reconstruction of the signal from its samples is possible when the power of frequencies above the Nyquist frequency is zero, a real anti-aliasing filter trades off between bandwidth and aliasing. A realizable anti-aliasing filter will typically either permit some aliasing to occur or else attenuate some in-band frequencies close to the Nyquist limit. For this reason, many practical systems sample higher than required to ensure that all frequencies of interest can be reconstructed, a practice called oversampling."

Parameters

The plugin which is responsible for setting different AA filters is called SettingsImageFilter. It has the following parameters:

on - Enable/disable image filter
type - 0 - Sinc, 1 - Lanczos, 2 - Catmull Rom, 3 - Triangle, 4 - Box, 5- Area
size - Image filter size (where applicable)

We'll cover the different types in the next section.

Avoid using filters with negative components (sharpening filters) like Catmull-Rom or Mitchell-Netravali when using the Progressive sampler. Doing so might dramatically increase render time as the sampler needs to take additional image samples to resolve the filter.

There are plugins for the different filter types - FilterArea, FilterBox, FilterCatmullRom, FilterCookVariable, FilterGaussian, FilterLanczos, FilterMitNet, FilterPoint, FilterSinc and FilterTriangle. You can see their parameters in the API docs.

Examples

Anti-aliasing Filters

Here is an example briefly demonstrating the effect of different anti-aliasing filters on the final result.

Note that rendering with a particular filter is not the same as rendering without a filter and then blurring the image in a post-processing program like Adobe Photoshop. Filters are applied on a sub-pixel level, over the individual sub-pixel samples. Therefore, applying the filter at render time produces a much more accurate and subtle result than applying it as a post effect. V-Ray can use all standard 3ds Max filters (with the exception of the Plate match filter) and produces similar results to the scanline renderer.

The Bucket image sampler was used for the images below.

Filter	Image	Zoomed-in image	Comments
Filtering is off			Applies an internal 1x1 pixel box filter.
Area filter, size 1.5 (default setting)			Slightly blurs the image, visually more pleasing than the box filter.
Area filter, size 4.0			More blurring.
Blend filter			Combination of a sharp and a soft filter, kind of dreamy effect.
Catmull-Rom			Edge-enhancing filter, often used for architectural visualizations. Note that edge enhancing can produce "moire" effects on detailed geometry.
Mitchell-Netravali			Allows control between edge-enhancement and blurring.
MItchell-Netravali, ringing=1.5			Strong edge-enhancement.
Mitchell-Netravali, ringing=2.0			Even more edge enhancement; kind of cartoon-style effect.
Soften			Gaussian blur.

Anti-aliasing Filters and Moire Effects

This example demonstrates the effect anti-aliasing filters have on moire effects in your images. Sharpening filters (Mitchell-Netralavli, Catmull-Rom) may enhance moire effects, even if your image sampling rate is very high. Blurring filters (Area, Quadratic, Cubic) reduce moire effects.

Note that moire effects are not necessarily a result of poor image sampling. In general, moire effects appear simply because the image is discretized into square pixels. As such, moire effects are inherent to digital images. The effect can be reduced through the usage of different anti-aliasing filters, but is not completely avoidable.

The scene is very simple: a sphere with a very fine checker map applied. The images were rendered with a very high sampling rate (15 subdivs, or 225 rays/pixel). This is enough to produce quite an accurate approximation to the pixel values. Note that the image looks quite different depending on the filter: