Wednesday, 19 March 2014

Sharpening for print

I started off by taking a portrait image which had been processed but not sharpened.

Using Photoshop and unsharp mask, I created 3 versions one close to the onscreen version, one fairly aggressive, and more aggressive. On screen, I thought I could tell a difference in the amount of noise visible in the background but this didn't transfer itself to the print. Using a magnifying glass and torch, there were a few areas such as the zip and pupils where I could tell that that sharpness had been applied.  I was not convinced that this had done the job I intended.


base
amount 90%
amount 150%
oversharp















I checked my camera, wondering perhaps if the settings in my camera were oversharp and maybe this was why my image had not sharpened visibly.

I took another image with more contrast and cropped to A4 using about a quarter of the photo. This time the exercise worked. 


base
amount 90%
amount 350% 
amount 250%
Summary of lily sharpening


Using a radius of 0.5 and a threshold of 1, I altered the amount of sharpening from the unsharpened image to 90%. There appeared very little difference on the monitor when viewed at 100%. In print and using a bright light and a magnifying glass, the difference is highlighted in red above. A few odd pixels had been brightened and in the middle there was a brighter edge along two of the stamens.

The effect of this surprised me. I hadn't really understood how unsharp mask worked and had been cautious in sharpening. The suggested starting threshold value is 90-120%. Gulbins and Steinmueller (2011) 

I kept the radius and threshold the same and altered the value to 250% to see the effect, illustrated by the blue line above. I could tell on the monitor that this was oversharp. A dark line was starting to become visible along the bottom stamen and the white petals had a darker patch on them (I wouldn't have called it a halo). 

I had still not seen the effects of noise, so kept the radius and threshold the same and increased the value to 350. On screen there was noise on the petals. This transferred to the print. The difference between 250 and 350 was alot more oversharpening and an increase in noise.

radius = "number of neighbouring pixels that are altered by the sharpening process" threshold = amount "of pixels to consider as an edge" 
amount = "how strongly edge contrast is increased" 
Gulbins and Steinmueller (2011) 

The exercise confirmed to me that at a threshold of 90%, radius 0.5 pixels and I would still have to be careful with areas which were too sharp, so continue to be cautious. There is an argument for only sharpening where it is needed. I felt it was a very beneficial exercise.

References
Gulbins and Steinmueller (2011) The digital Photography Workflow Handbook, Rockynook, C.A p112

No comments:

Post a Comment