Pre and Post resize sharpening


(Harshula) #1


I’ve read both:

Is it desirable to have pre and post resize sharpening of the same image?


(Ronald E Chambers) #2

Personally, I would only sharpen as the last process before display of the image.

(Morgan Hardwood) #3

@hash I stopped using pre-sharpening ever since RT got the post-resize sharpening feature.

(Alberto) #4

same here. I only sharpen when resizing. there are other tools that work better for me instead of pre-sharpening, especially contrast by detail level


There’s certainly nothing wrong with applying pre and post sharpening.

Many people follow the three stage sharpening philosophy
The key here is that the first pass is to compensate for softness caused by the kit, e.g. AA filter, soft lens etc. It is important to keep this stage gentle and do not introduce artefacts like halos.

I use deconvolution for pre sharpening (warning this doesn’t work well for noisy images)
radius 0.67
amount 67
damping 20
iterations 100
and then unsharp mask with default settings for post resize sharpening (stage 3).

RT only has two sharpening stages, but any of the contrast tools (CBDL, edges, micro contrast, wavelet tools etc) can also be considered sharpening, because they increase apparent sharpness, and I would consider these as stage 2 creative sharpening.

As an aside @agriggio how easy would it be to allow for deconvolution and unsharp mask in the main tools, rather than having to select one or the other? I ask you because a) you’ve participated in this thread and b) I notice you’re super keen and active in development (many thanks btw) :slight_smile:

(Alberto) #6

hello @james

Technically it would be quite easy, it’s just a matter of implementing the new UI. As to whether this feature request would be accepted, well, it’s not my call so I don’t know :slight_smile:

well, thank you :blush: but I mostly do it for myself, I’m an avid RT user and if I see some rough edges I try to help in smoothing them.


@agriggio Thanks, I’ll give it some more thought as to whether to make a request on github or not. I think there’s less of a need now there is post resize sharpen.

(Glenn Butcher) #8

For my camera and shooting, I’ve only found utility for sharpening after reduction resizing. Even the good interpolation algorithms (lanczos3, catmull-rom) still take some perceived acutance out of the image, and for my money a simple USM min-strength sharpening right after resize produces the best result. And really, going from no-sharpen to the first setting produces the most benefit; after that, halos start to appear.

I’ve tried various sharpening right out of raw conversion, and I just don’t see any benefit viewing the overall image for the time it takes. Probably a different story printing large…

(Desmis) #9

You can also use “wavelet”, with “Edge sharpness”, but by activating “edge detection” and if you want “enhanced algorithm”.
In this case, you do not only have, as in “CBDL” or “wavelet contrast”, an increase of both the edges and the microcontrast, but a real “sharpening” which gives results “close” to those of “RLdeconvolution” or “Unsharp mask”, but different.

I innovated this approach using several concepts:

  • Wavelet decomposition which allows to choose only the desired level of detail
  • Canny algorithm adapted to wavelets
  • High pass filter on the 3 components “vertical - horizontal - diagonal” to detect the edges
  • Low pass filter to take noise into account and eliminate details to too low gradient
  • Etc.

Wavelet is at the end of the process, just before CIECAM

You can consult Rawpedia (in french)


@ggbutcher Hi Glenn, yes you are right. For downsizing stage 1 is pointless.

@jdc Thanks Jacques, I have only played a little with wavelets. I will try the steps for “real sharpening”. (and thanks for all your algorithms you introduce to RT too :slight_smile: )

(Harshula) #11

@james Hi James, Can you please clarify whether you use sharpening before and after downsizing on the same image?


Hi @harshula Yes I do.

and I’m happy with the results, so I probably won’t change what I’m doing, because it doesn’t cause me any problems.

However, as part of this discussion I have been persuaded that for downsizing, post resize sharpening is enough on its own. So if you are just beginning to develop your own workflow. I think you should try just post resize sharpening and see if you like the results. However, if you don’t like them, I will stick by my original line [quote=“james, post:5, topic:3614”]
There’s certainly nothing wrong with applying pre and post sharpening.

(Ronald E Chambers) #13

I would say a deconvolution process is fine because it is global so that local features are not changed as compared to any type sharpening which is local process and will change within a few pixels. Sharpening must be last because of that. The finer the features the better the chance to damage them with sharpening .