|
Banding in low colour resolution workflows |
|
|
|
Contributed by Kai-Uwe Behrmann
|
|
Monday, 26 September 2005 |
Banding is in low colour resolution imaging workflows a common phenomenon. Contrary the artikel tries to show that banding, using high colour resolutions, is unlikly to happen. The high bit depth capable open source application CinePaint is used for the demonstration of banding. Consider this artikel as well as an tutorial and try yourself.Step by step:
- open an new image in CinePaint - from the top menu choose: File −> New and create a new image, not too tall (1000 or better 400 pixel wide)
- set the bit depth to test - from the image menu choose: Image −> one of:
- 8-bit Unsigned Integer
- the basic bit depth used in the web
- 16-bit Unsigned Integer
- used for high colour resolution scans and camera raw images
- 16-bit RnH Short Float
- lower colour resolution HDR
- 32-bit IEEE Float
- high colour resolution HDR
- from the CinePaint toolbox choose the gradient tool and create an gradient over the whole image you created befor
- now we manipulate the image very heavy to make any banding clearly visible - from the image menu choose: Image −> Colors −> Levels
- move with the mouse the black (left hand) and white (right hand) triangels, in the Levels window below the histogram and the first gradient gradient, close to each other; see here
- watch the resulting manipulated gradients of each colour resolution and compare here
an gradient before manipulation
Levels dialog before manipulation:
and after:
Comparision
After manipulation the various gradients could look like below:
8-bit Unsigned Integer
<-->
16-bit Unsigned Integer
16-bit RnH Short Float
32-bit IEEE Float
As a alternateive comparision see drawing in CinePaint:
Conclusion
Tradition shows 8-bit LDR could serve the needs of the printing industry in the last years. Today, where technical equipement has improoved much and in the same time the average technical knowledge has decreased, we enface a new situation. 16-bit is now standard for photographers. Working in HighDynamicRange (HDR) is a new need. IEEE 32-bit float gives, beside an increased precission of 22-bit per channel, the freedom of manipulation for HDR content,
Recomendation
My suggestion is to use low colour resolution images (8-bit and 16-bit float) for the web and similiar purposes, where bandwidth is limited.
For manipulation and archiving use eighter 16-bit LowDynamicRange or 32-bit IEEE float (HighDynamicRange).
The LowDynamicRange formats like 8-and 16-bit are useful for scans and camera shots. For bracketing exposures or HDR camera output use a float format (HighDynamicRange).
Terms
- LowDynamicRange - LDR
- the dynamic range ends with the maximum amount being the same for white. (jpeg, png, integer-tiff, bmp)
- HighDynamicRange - HDR
- the format is able to store intensities not printable on paper, like specular highlights (cineon, float-tiff, openexr)
Kai-Uwe Behrmann
[28. September 2006 : Version 1.1]
[26. September 2005 : Version 1.0]
|
|
Last Updated ( Thursday, 22 February 2007 )
|
Home
