Image Analyzer examples
Image Analyzer examples
The Increase Local Contrast dialog has grown into a multifunctional control with possibilities to adjust both local
and global contrast, tone mapping, color saturation, white balance, noise reduction and sharpening. Most of the
parameters do, however, still control different aspects of local contrast/detail. The rest of the adjustment
possibilities were included here for easy access.
The basic color image format in Image Analyzer is 8-bit RGB, but everything under this dialog woks in 32bit precision
so no color tones are lost between the different filters involved. If you use File|Import for HDR to open an
image file you can also open a 16-bit file and work with it directly in Increase Local Contrast without loss of
precision.
We will here describe the parameters one at a time.
Increase local contrast can improve photos that are underexposed or has uneven light. It can also give a dramatic effect
to otherwise dull pictures. Decreasing the filter size will increase the contrast locality and increasing it will make
it work on a more global scale. If the picture is very under exposed then it may amplify any noise present, especially with
small filter and high maximum contrast settings, so it can be a good idea to also apply noise reduction.
Basically the filter stretches the histogram to utilise the full possible value range. In an 8-bit picture pixels can
values between 0 and 255. If in some area of the picture the pixels have values between 20 and 150 they will be scaled to
span the full 0-255 range. The Contrast neighbourhood parameter determines how big an area around each pixel is
evaluated in order to determine how much its value should be stretched.
Example of lowering the Contrast neighbourhood and increasing the Outlier reduction:
(Hover the mouse over the first image to see the result change in-place.)
Determines how much the pixel value range can be stretched. 100 means that no matter how small the input value range is it
will be scaled to the full 0-255 range.
If you see false colors in some areas when decreasing the Contrast neighbourhood parameter you can decrease the
Maximum contrast in order to avoid discoloration and still have a high contrast localization.
Usually you can just leave this at it default value 50.
Increasing the Outlier reduction will allow stretching the histogram so that a few values may fall outside the 0-255 range and thus be clipped to 0 or 255. This will disregard noisy outlier pixels in the local contrast filtering and can also be used to get extra contrast in the result.
This can be used to achieve more contrast localization without getting halos near edges. Often you will get better quality setting the Tone mapping parameter below 1 though.
This is plain global contrast.
Global color saturation. Sometimes local contrast will also increase the overall color saturation, so this parameter can be
used to achieve more natural looking colors.
You can also move it all the way to 0 to make a color image monochrome. If you use this to make mono images you will often
get more striking results by decreasing the Contrast neighbourhood and increasing the Outlier reduction.
This usually give much more interesting monochromes than just removing the color.
When set to a value below 1 it will increase the local detail level. This can be used to make pictures look more rough.
When set above one it works as an edge-preserving smoothing filter.
The effect of detail enhancement (photo from Toul Sleng prison, Cambodia, � Michael Vinther 2012):
Tone mapping is frequently used with HDR but can also be useful on single images.
Setting the value below 1 will increase the tone localization.
This parameter has a similar effect to Contrast neighbourhood, but is more targeted at HDR than just local exposure
correction. The algorithm also requires somewhat more processing power.
Detail enhancement and tone mapping (photo from Mostar, Bosnia-Herzegovina, � Michael Vinther 2015):
A threshold for what is considered details in Detail handling and Tone mapping. Usually this should be left at the default value, or you can try lowering it to enhance the effect.
By default Band pass min is set to the lowest possible value and Band pass max is set to the highest. This will make Detail handling and Tone mapping work evenly over the whole image. You can change the values to limit the affected frequency range. For example, detail enhancement might enhance noise and this can be avoided by increasing Band pass min. On the other hand, limiting it to only working on the low frequencies can also give the image a more rough look without changing the overall lightning.
Lighten dark areas.
Darken light areas.
Edge-preserving noise reduction filter strength.
This is the same filter that you can access under Operations | Filters | Local stats denoising filter.
Noise reduction example: (Hover the mouse over the first image to see the result change in-place.)
Additional color noise reduction.
Additional noise reduction around edges.
Spread of the underlying noise reduction Gauss kernel.
Edge sharpening filter strength.
This can be increase to avoid enhancing the noise when sharpening, but at the cost of less detail sharpening.
Determines which frequencies are amplified when sharpening. Increasing the value will make the sharpening filter boost larger features.
If you make a selection in an area of the picture that is supposed to be white (or neutral gray) before opening the dialog,
this slider will allow you do white balance correction based on that area.
You can also press the Pick button and then click with the middle mouse button in a white (or neutral gray)
area.
Note that the area you choose should not be overexposed because then the correction will not be correct.
White balance correction example based on a selection in the white jacket, value=0,58. (photo from Marrakesh, � Michael Vinther 2015):
Approximate color temperature adjustment. The values in the scale assumes a color temperature of 6550K in the original image. Moving the slider to the left will produce more reddish colors and moving it to the right will produce more bluish colors.
Color plane weights in saturation adjustment. These are particularly useful when creating monochrome images from color images.
Try to set the Color saturation to 0 and see how adjusting the R/G/B weights affects the result.
In the following color to monochrome example reducing the weight of the red channel enhances the details in the face because
it is slightly over exposed in the color version.
Convertion with standard R/G/B weights (first) and with reduced red weight (second). Hover the mouse over the first image to see the result change in-place.
