Edge Aware Filter

Noise removal is an essential part of compositing virtual elements onto live footage. The Edge Aware Filter provides two methods of smoothing noise in an image, while maintaining detail along along the edge of an object.

Overview

The Edge Aware Filter has two options for removing noise while preserving edge detail. The first is a bilateral filter, which is essentially a Gaussian blur produced by weighting the contribution of neighboring points by there distance and color difference. The bilateral filter is very effective at smoothing areas with mild to moderate noise. The second filter is a median filter, which scans a region around the current pixel and sets the color value to the median of the samples. The median filter is very effective at removing impulse, or salt and pepper noise.

Inputs

Results

A simple outdoor image. Noise is very visible in the sky.

Output from the bilateral filter with a blur size of 1.0 and a color blend size of 0.1.

Output from the median filter with a blur size of 0.3.