A study of sharpening methods using the K-7

Started May 8, 2011 | Discussions thread
ForumParentFirstPreviousNextNext unread
Flat view
labnut
Senior MemberPosts: 1,016
Like?
A study of sharpening methods using the K-7
May 8, 2011

Here I compare in-camera sharpening (K-7) with various sharpening methods available in Gimp.

In my previous post I measured the effect of the various in-camera sharpening methods and made the observation that their quality seemed high, now I back up that statement.

http://forums.dpreview.com/forums/read.asp?forum=1036&message=38376244

First though, a note about what I did.

I photographed a back lit 5 degree slanted edge in RAW, using a Sigma 50mm macro lens on a K-7. This one image was sharpened by all the methods shown below in the diagram. I assessed the effect of the sharpening using four measures (and the tool QuickMTF):

1. edge sharpness as the degree of blurring of the edge in pixels.
2. noise as standard deviation of highlight pixels.
3. overshoot height as percentage of brightness.
4. overshoot width in pixels.

These four measures completely characterise the effect of the various sharpening methods.

The results are shown below (smaller edge blur is better) and compared with an un-sharpened image, at the bottom. All sharpening methods used their default parameters. My comments are below the graph.

Notes.

Sharpening methods work by finding edges in the image and reducing the blurring of the edges. For example, in my tests the edge originally had a blur of 1.97 pixel and this was reduced to 1.25 pixel by the Smart Sharpening algorithm, which is nearly the very best result that one can get.

But sharpening introduces unwanted effects:

  • it multiplies noise already in the image

  • it introduces halos around image outlines that become visible if the overshoot height or width are too high.

In the graph above it can be seen that Smart Sharpening is very effective but it has a high cost in terms of noise and overshoot. Unsharp mask is similar but the halos become more visible because the overshoot width is even greater.

Refocus is also very effective, has very little overshoot, therefore no visible halos, but you pay a higher price in noise.

The in-camera FS2+4 setting would seem to be the best all round choice. Sharpness is very good, there is only a slight noise increase while overshoot height and width are also mild.

A good strategy for choosing a sharpening method is to start with image noise. If this is already low you can choose one of the methods that multiplies noise and see no undue side effects. The Refocus algorithm might be appropriate.

If there is already image noise then avoid the Refocus methods.

In general, Pentax's in-camera sharpening methods compare very well with other popular algorithms.

ForumParentFirstPreviousNextNext unread
Flat view
ForumParentFirstPreviousNextNext unread
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow