CM Week 4 - Measuring Statistics : Some Questions

[dkrassen]

Hello,

From the information you provided, I see much use for Delta E in determining how close two colors are to each other.  I understand a dE of 3.0 is considered to be undetectable by the human eye.  Would you elaborate on what use other values for dE would have in correcting photos?  For example if a dE of 3.0 is undetectable by the human eye - is a dE of 4 so close to 3 that it also means the two compared colors are so close that is considered a good result? 

Please elaborate on what the signal to noise (sn) actually means and how to use it?

Finally, you state that samples can be saved to a text file and imported in Excel for statistical analysis; please elaborate on how this would be performed?

Thank you
David

[ggroess]

David,

You have hit upon the one weakness I have in the Whole program...These are Questions for Mike...If you can wait for next week he will be back and can better answer you than I can.  If any of the Power Users are watching me flounder...Right about now is when you should jump in with your 2 cents worth of information....

I'll forward your question to mike...

[-default]

These are both fairly arcane functions, but they do have a number of  verypractical uses.

Delta E aka dE

---

Delta E was actually added as a custom feature, due to a customer request several years ago.  I added in the standard deviation function because I thought it might be useful as well.  The feature has been carried over to the main product, and made available for everyone.  Here are some practical uses for the two rather arcane functions.

Delta E is, as you already know, a function giving the approximate visual difference between colors.  A dE of 3 is considered by the industry to mean that the two colors in question match visually.  Here are some practical uses.

1) Compare two camera bodies: photograph a chart with camera body 1, swap in camera body 2, and photograph the same chart. again.  In Photoshop, load the first image, Alt click each of the patches to create a set of samples, and save them.  Load the image from body 2, load the samples, select all the samples and enable delta E display - voila you have a deltaE for each of the patches.  Save the samples to another file, and open as a csv in Excel.
2) compare prints to a proof.  Scan the proof, create sample points as above, select all, and save the samples as your reference file.  Scan each print, load the sample points, and read off the delta E values as with example 1.
3) compare two areas of the same image - alt-click image point #1, lock it, and drag the point to image point #2.  This will display the deltaE of the two locations on the same image.
4) deH (Delta E Hue) - like the previous example, but set a sample point to brightness mode - move the point and the brightness will change, and the resulting deltaE will be a readout of the hue difference only.

Standard Deviation aka sd

---

standard deviation is a measure of how noisy the sample area is.  It is the root mean square difference between each channel in a sample and the average value of the channel values.  The sample area in question is the square area used by the sample.

1) find the noisiest channel in an image by looking for the highest sd value.
2) compare the amount of noise in images from two or more cameras.
4) match the appearance of a cloned area, or other manipulate area, to the original image
5) there are a number of arcane optical calculations that are based on measuring sd.  One such is the Modulation Transfer Function, which uses sd to characterize the sharpness of an image by using a test target comprised of alternating black and white bands or other pattern, and calculating the standard deviation of the patterned areas versus pure black and white areas.  this is a very sensitive and reliable way to characterize the resolution of an optical system.

As with delta E, saving a set of samples with sd enabled will create a tab delimited csv file, suitable for importing into a column based program such as excel.