AS Roughness Histogram and Chroma Noise

Moderator: jsachs

Post Reply
tomczak
Posts: 1367
Joined: April 25th, 2009, 12:56 am
What is the make/model of your primary camera?: Fuji X-E2
Contact:

AS Roughness Histogram and Chroma Noise

Post by tomczak »

Am I correct in thinking that Roughness histogram in AS is calculated based on HSV-V values, regardless if subsequent blurring is done to all channels or chroma only? In other words if one blurs chroma only (however chroma is defined), the thresholds will protect brightness (i.e. HSV-V) edges, not necessarily chroma edges, even though chroma is being blurred and brightness left alone?

I did some more experiments on chroma noise removal. As Jonathan mentioned in another thread, the difficulty is in removing low frequency colour blotches: the blurring radius and amount need to be fairly large, otherwise the low frequency noise is replaced with even lower frequency, uglier patchy mottle.

Such aggressive blurring also dilutes legitimate colour edges. The (upper) threshold slider under the roughness histogram can help to protect such edges from being blurred, but...

1) as it is, AS doesn't separate chroma from luminance well, and blurring chroma-only leaves brightness poke marks that don't look too well.

2) but if an image is disassembled by hand prior to AS (e.g. into Y'CbCr components or, somewhat less distinctly, into HSL components), and then one or two chroma channels are blurred (Cb seems to be particularly associated with chroma noise), chroma edges that shouldn't be blurred can be effectively protected with AS thresholds. But those thresholds work by being calculated and set based on the actual roughness histogram of chroma channel being blurred, not on the brightness channel of the entire image (if how I think AS works is actually true).

I can't be sure, but it appears to me that blurring chroma, that is well separated from luminance, with large radii and amounts but with proper threshold based on roughness of the chroma channel being blurred, can effectively segment the chroma-noisy image into more or less contiguous regions of constant colors and protect sharp transitions between them, while still removing low frequency colour noise in their midsts.
Maciej Tomczak
Phototramp.com
Post Reply