Smart Object - Stack Modes
With Photoshop, you can combine a number of layers into a single stack as a Smart Object. With a Smart
Object Stack, you can determine how the pixels in each layer interact with
the pixel in other layers within the stack.
The images in the stack are usually a series of photos using a tripod and
the same exposure settings. Take a number of photos of a static object
with moving objects (people or cars, for example) over a reasonably short
period of time (minutes, not hours, so the lighting doesn’t change much).
Select several related or contrasting images and add them as layers to a single PS file, select all layers:
1. Create a Smart Object by choosing Layer→Smart Objects→Convert to Smart Object.
Note: If the photos were not shot on a tripod, “Align Layers” before creating Smart Object Stack,
2. Return to the Layer→Smart Object submenu and select the Stack Modes submenu that’s now
available.
Below are the options that determine how the content on the layers within the Smart Object interact to
produce the appearance of the Smart Object itself.
• Entropy
o The binary entropy (or zero order entropy “randomness”) defines a lower
bound on how many bits would be necessary to losslessly encode the
information in a set.
o entropy = – sum( (probability of value) * log2( probability of value) )
o Probability of value = (number of occurrences of value) / (total number of
non-transparent pixels)
• Kurtosis
o A measure of peakedness or flatness compared to a normal distribution.
The kurtosis for a standard normal distribution is 3.0. Kurtosis greater than
3 indicates a peaked distribution, and kurtosis less than 3 indicates a flat
distribution (compared to a normal distribution).
o kurtosis = ( sum( (value – mean)4 ) over non-transparent pixels ) / ( (
number of non-transparent pixels – 1 ) * (standard deviation)4 ).
o
• Maximum
o The maximum channel values for all non-transparent pixels
Bob DiNatale Version 2.0 Page 1 of 3
• Mean [Effective for noise reduction]
o The mean channel values for all non-transparent pixels
What will this do our Smart Object? It will take the mean value of each
pixel. As a result, our images will look cleaner, you will see an
improvement in the noise and will obtain more details.
• Median [Effective for removal of unwanted content from the image]
o The median channel values for all non-transparent pixels
• Minimum
o The minimum channel values for all non-transparent pixels
• Range
o Maximum minus the minimum of the non-transparent pixel values
Bob DiNatale Version 2.0 Page 2 of 3
• Skewness
o This is a measure of symmetry or asymmetry around the statistical mean
o skewness = (sum( (value – mean)3) over non-transparent pixels ) / ( (
number of non-transparent pixels – 1 ) * (standard deviation)3 )
• Standard Deviation
o standard deviation = Square Root(variance)
• Summation
o The sum channel values for all non-transparent pixels
• Variance
o variance = (sum( (value-mean)2 ) over non-transparent pixels ) / ( number of
non-transparent pixels – 1)
Bob DiNatale Version 2.0 Page 3 of 3