Can someone explain sharpening?

[JDArt]

Can anyone explain from a technical standpoint what "sharpening" photos in software actually does to a digital image?

[Michele]

Good question.... I definitely don't have a clue on how to answer that one... I am interested in the answer as well...

[ohenry]

First, sharpness and sharpening are two separate terms. Sharpness is the result of focus, exposure, lens quality, film/sensor, camera movement, etc.

Sharpening is a process applied to digital images, whether scanned from a slide, print, or direct image from a digital camera. It is a process of increasing the contrast of the pixels where there is an edge or natural difference in contrast, brightness, or color. Sharpening is a "destructive" process in that it is applied directly to the image pixels and permanently alters pixel contrast values. Too much, or repeated, sharpening can and will degrade the image.

There are many ways to sharpen an image, some better than others for given circumstances and each having their own merits. Digital cameras apply a given amount of sharpening, which in most cameras can be turned up or down or even off. I prefer to shoot RAW and never sharpen from the camera. Often when you see someone say that their XYZ consumer point and shoot gives better images than their $1500 dSLR, it's because of the amount of sharpening (and other parameters) applied by the camera. Post sharpening is generally preferred if you plan on doing post production work. Because of the destructive process of sharpening, it is generally applied after tonal adjustments are made. Bruce Fraser feels that sharpening should be applied in a 3 step approach. First for the softness of the capture device, then again for creative sharpening applied to specific areas of concern, and finally sharpening that is targeted to the final output. Personally, I follow his method and use his sharpening software 99% of the time that I sharpen.

Unsharp Mask is probably the most popular method of sharpening. In Photoshop (not sure of other programs), you can control the amount of sharpening, which increases the contrast of pixels by either lightening or darkening them. If too high, you'll see halos or fringing at the edges. Radius is another control available for USM. This setting determines the width of the sharpening halos. Obviously, the wider the width, the more pronounced the effect. If too wide, the image will appear oversharpened. Radius and amount work together back and forth...when you increase one, you should decrease the other. Threshold is the final setting control. It controls which pixels the USM filter will consider. It has a value of 0-255, so if you enter a value of 8, the filter will sharpen a pixel only if there's a difference of at least 8 tonal levels between it and the pixel next to it. Typical threshold values range from 0.5 to 8. I'm not really sure why threshold goes as high as it does as higher levels rarely will help the image. I've seen many comment that good starting points are: Amount of 300 then use radius and threshold to back off the intensity until you reach something that pleases your eye.

Other methods of sharpening include edge sharpening, high pass filter sharpening, and, of course, plug ins such as Fred Miranda's, Pixel Genius, and others.

The amount of sharpening you apply depends on the size of the image and the output medium. What looks good on your monitor may not do well on an inkjet printer and what works for your inkjet printer may not work for commercial print machines.

Hope I answered your question somewhat. Obviously, this is a much more in-depth topic than I've outlined briefly here.

[Mr. Pickles]

OHenry pretty much has it. It is a destructive means of increasing the "line" between things. These things are usually seperated by colors. Your ability to sharpen well will be seen in these color differences. Lots of times, you can see images that have a "halo" around them. They look really sharp, but if you look, they look fake too.

In some respects, sharpening with software is like tell a 16 million color image to be a 32,000 color one. The computer looks at each pixel and its neighbors. If it sees a "match" it changes it. It is NOT a good thing, but it is a necessity.

[scoundrel1728]

The term "unsharp masking" is a legacy from the days before digital photography. Back in those days, a slightly blurred, low-contrast contact positive - the unsharp mask - was made from the original negative. The two pieces of film were then sandwiched together and printed out. High-contrast paper was used to bring the overall contrast back up to its original value, with the edges then developing especially high contrast because these areas were not partially masked by the positive.

A similar effect is applied digitally today, where the radius setting corresponds to the amount of blur and the intensity to the amount of contrast enhancement. Because unsharp masking also tends to emphasize noise, as high-pass filters generally do, the threshold is added to counteract this effect while still enhancing the contrast of the real edges.

The effects of denoising and sharpening are largely complementary. Things done to increase sharpness also tend to increase perceived noise and vice versa. It can be a real balancing act to sharpen a noisy image or to denoise an image without undue loss of detail.

[mikej]

It's an optical illusion.




Mike

[dc2]

Mike has a short answer but accurate

[scoundrel1728]

I used to think so too but now I'm not so sure...

[blumesan]

For an excellent essay on the technical aspects as well as the practical application of sharpening go here

http://www.thelightsrightstudio.com/...ningSkills.pdf

Cheers,
Mike

[dc2]

Originally Posted by scoundrel1728

I used to think so too but now I'm not so sure...

I am not sure why your not sure for sure

[scoundrel1728]

Originally Posted by dc2

I am not sure why your not sure for sure [why "sharpening" is an optical illusion that does not actually restore detail]

Ideally, and if carried out perfectly, the mathematical process of deconvolution would be able to undo the bad effects of camera/subject movement, lens aberrations, diffraction, and other effects that would degrade the image. In practice, the presence of noise, lack of exact knowledge of the exact deconvolution function, and the fact that deconvolution can involve division by zero or nearly zero all prevent the reversal from being carried out perfectly. Nevertheless, even an approximate reversal can improve the resolution of the image. Unsharp masking mimics in a crude way the result of deconvolving certain kinds of image degradation. For those familiar with electronics, the result is much like Q-peaking a circuit to improve its high frequency response.

[dc2]

Originally Posted by scoundrel1728

Ideally, and if carried out perfectly, the mathematical process of deconvolution would be able to undo the bad effects of camera/subject movement, lens aberrations, diffraction, and other effects that would degrade the image. In practice, the presence of noise, lack of exact knowledge of the exact deconvolution function, and the fact that deconvolution can involve division by zero or nearly zero all prevent the reversal from being carried out perfectly. Nevertheless, even an approximate reversal can improve the resolution of the image. Unsharp masking mimics in a crude way the result of deconvolving certain kinds of image degradation. For those familiar with electronics, the result is much like Q-peaking a circuit to improve its high frequency response.

Funny I mentioned a guy I saw who had a 1ds2 2 in another post, he said he did not sharpen any of his pics but instead relied on top end prime lenses to do the job for him . I was very suspect but some prints 13x19 he showed were very impressive some of his reasoning was very similar to your thoughts on this

[scoundrel1728]

Not sure how that has anything to do with deconvolution though, since the ideal image didn't get convoluted with a big point-spread function in the first place. Therefore, his images were sharp to begin with and needed little if any after-the-fact sharpening.

[Mr. Pickles]

Maybe the guy shot just JPG

[dc2]

Originally Posted by Mr. Pickles

Maybe the guy shot just JPG

No the guy was shooting in raw

[Mr. Pickles]

That guy shot in RAW and didn't sharpen at all? Dang. Don't tell him about our Bi-Weekly Photo Contest, okay

[dc2]

Originally Posted by Mr. Pickles

That guy shot in RAW and didn't sharpen at all? Dang. Don't tell him about our Bi-Weekly Photo Contest, okay

I wont* but I wonder if he will lend me his lenses
I personally sharpen only after the raw process . I only have one prime lens which is very sharp so I must try and see if his idea is as good he believes I will post the result and my thoughts on this.