Sharpening, how do you do it?

[rosko]

Interesting thread.

Most of us (including myself) have the (bad) tendency to sharpen our images, sometime  without discernment, often systematically.

I try to moderate this habit, asking this question : '' Does this image really need to be sharpened'' ?.

I reckon this method it's fine if you want to save a valuable picture. in this case I will try to find the right balance between details an grain.

In any case, the higher are iso, the more destructive the sharpening is.

I've noticed that some settings (using View NX) like contrast or shadow protection add grain.

If anybody knows the best way to sharpen without grain. please, let me know...

[Bjørn Rørslett]

John Geerts: Sharpening of RAW is a non-concept. It cannot exist by definition.

Rosko: explore the potential of layering or brushing in detail with the history brush (PS). You can protect part(s) of an image by making masks. And so on ad infinitum.

[rosko]

Thank you Bjørn. I will dig your advice with Gimp, hoping there are the same features as PS (which I haven't used yet) I am really weak in processing : I don't even know how to create layers with Gimp...

By the way, the example given on Seapy' link shows obvious improvement in sharpness, but the bird is over saturated to my taste. The original pastel colors are so much nicer.

[John Geerts]

John Geerts: Sharpening of RAW is a non-concept. It cannot exist by definition.

Yes. What I meant was when shooting in jpg there may be a sharpening in-camera.

[Bjørn Rørslett]

Yes, with jpg you can select in-camera sharpening. This might even be a camera default setting. I suggest it be turned off immediately. In-camera sharpening can only cause issues later in the work flow.

For sharpening or any further processing of a jpg, start by making the imported file 16-bit and scale it 4X (even bigger if you need to deliver a huge print). Set the colour space to the largest your software can handle. You might add a layer copy slightly blurred and merge the two. Then work on that file as usual.  When ready, downscale to the final size and do the sharpening there. Convert to the wanted output colour space and 8-bit.

[Tristin]

Bjørn, what's the reason for sharpening after downscaling?

[elsa hoffmann]

I rarely sharpen any of my photos - if I do - its on wildlife but even then... I forget
I dont want any more sharpness on ppl especially

(D800 & 135 f2 DC) shot at f8 with studio lights

I photographed 11 managers of a local chain store - and the images printed beautifully sharp

[Fons Baerken]

In photoshop there is the option to apply bicubic sharpening when resizing.
If you want to publish an image for web, for instance where the original size is 4900x3200
downsize in half steps not straight to 1024x800, and while doing the image will smart sharpen
and so on.

[armando_m]

Whatever I do with sharpening, it always looks overdone when seen later, so lately  I tend to to reduce it to 50% of my initial attempt, and then things do look better

[Bjørn Rørslett]

Bjørn, what's the reason for sharpening after downscaling?

Because this is the final image ready for use.

[Tristin]

Thanks Bjørn, wanted to make sure there wasn't a technical point I didn't see.

[Peter Forsell]

Bjorn already mentioned the technical point in an earlier post. Any geometric translation (resize, rotate, distortion correction etc. etc.) will wreak havoc if the image is already sharpened. Best to do sharpening at the latest possible moment.

[Akira]

Tristin, thanks for starting this thred for the seemingly entry-level but actually very essential topic!

[simsurace]

For Bayer sensors, where the image after raw conversion is obtained by interpolation, slight capture sharpening is simply an extension of the interpolation algorithm. Depending on the Raw converter, setting all sharpening controls to zero results in an incredibly soft image, particularly for sensors which have an optical low-pass filter. I don't know what would be a neutral setting in that case.

However, sharpening to the extent that it is visible at the final size should be reserved to after down-sampling. This is for several reasons: 1) it is hard to predict how sharp the result will be when sharpening earlier, 2) the downsampling algorithm itself can be tuned to give sharper or less sharp results and therefore interferes with the manual sharpening, 3) the artifacts that arise because of the sharpening will be combined with downsampling artifacts to yield even more artifacts, 4) (related to 1) the scale (radius)at which sharpening is applied depens on the downsampling factor, which is not known beforehand, and if it is, it is still better to sharpen after.

I'm sure you can think of more reasons.

[PeterN]

Thanks for the informative post.