If I count peaks along a segmented line in ImageJ and the the numbers are the same between two images but the plots show different results, what does that say about visual assessments about AFM images vs those which are digitally manipulated.

I understand the sensitivity of programs, some plots show every pixel, very noisy, but is this relevant information. It is possible to smooth out that noise, it is done in many different applications all the time. Is smoothing the noise helpful.  Here are two images, plotted in ImageJ, the one on the left enhanced with a color height map, the one on the right proccessed many times in three programs. Both image however present to the “eye” seven peaks.  The image on the right was processed thusly:

Original color-height map image where distortion was corrected in photoshop with edit>transform>skew, exported as a tiff, imported into gwyddion where an integral transform>CWT-5px-gaussian was performed, then in photoshop converted to grayscale, contrast enhanced and an additional 10px gaussian filter was applied as well as an unsharp mask at 50% amount, 50px radius, level of threshold=0.

What changes in images from one computer to the next mean for image analysis – have you considered this.  I did not, and have been doing image analysis for decades, literally, from the grid in the occular (as it was the only tool available 55 years ago) to using dedicated imaging programs (ImageJ, Gwyddion, CorelDRAW, Photoshop).

Over that time I have a reasonably good idea that if i can see a change in structures, positions, associations (in organelles), numerical densities, size, etc that I can do morphometry and there will be significant (p=0.05) change in what I am measuring.  I am not claiming to be clairvoyant, and underscore that I photograph the images in a random stratified manner and keep the image coding unknown until the data are analyzed.

So when i saw a relatively highly processed image of DMBT1 ring molecule change visibly during the save, import to a remote computer, and opening in ImageJ vs the image being saved in photoshop I began to worry that maybe something was going on that i did not recognize….  literally hundreds of other images have been moved from one computer to the remote computer without any such change….  why? just this image, why? not detected before?

Photoshop screen print (lower right) and photoshop saved (lower left) import from coreldraw (upper right). Each is different.

Here are two LUT plots of an arm of DMBT1 which were taken with different imports and screen prints of the same image. THey are close but not identical. I have hundreds of plots of similar arms, none is perfect, no image processing is the best, repeating the measures in a variety of formats is perhaps the best way to diminish the various differences in how programs handle images.