Filter Forge 2.0 finally has these essential components everybody has been asking for – you can now perform minimum, median, maximum and custom-percentile filtering. While the components don't support HDR colors (because their algorithms rely on a finite dynamic range), they can handle floating-point colors and are able to optionally anti-alias their output, which is unique to Filter Forge's implementation of these algorithms – we haven't seen that in other image-processing apps. Like other bitmap-based components in Filter Forge, these components can efficiently handle any image sizes up to 65000x65000 pixels.
The Minimum component replaces every pixel of the source image with the darkest color of its surrounding pixels within a certain radius, which has the effect of spreading out dark areas and shrinking bright areas:
The Median component replaces every pixel of the source image with a median color of its surrounding pixels within an adjustable-radius neighborhood. While the primary use of the median filtering has historically been noise removal, this component is very useful for creating a wide variety of artistic effects thanks to its ability to "simplify" the source image by removing small details while preserving the edges:
The Maximum component replaces every pixel of the source image with the brightest color of its surrounding pixels within a certain radius, which has the effect of shrinking dark areas and spreading out bright areas:
Basically, the Percentile component is a generalization of the other three components – it can do everything that Minimum, Median or Maximum components can do. When you set its Percetile input to 0, you get the same result as Minimum, 50 yields the same result as Median, and 100 is identical to Maximum – but you can just as easily set Percentile to any value you want! The adjustable percentile is very handy for artistic purposes – for example, the 95th percentile produces results that are visually close to those of Maximum (see the above picture), but without the ugly squares:
What's more interesting is that the Percentile input can be mapped with images. In the example below, Percentile is mapped with a gradient, so at the left edge of the image the component behaves as Minimum, then it gradually moves to Median behavior in the middle, and finally to Maximum at the right edge:
Class is going REALLY well! They are loving your stuff.
(after showing Filter Forge to a class at Adobe MAX 2009)
Senior Creative Director
“I do think you guys are likely the best software company I've ever had dealings with.”
“I'm thinking Filter Forge should come with a warning: 'This Product could be Habit Forming'.”