Is there (or will be) a universal method to process AFM images?
There is a major difference between AFM images and photographs. Whereas both are made of pixels, raw AFM images contain real numerical values (32 bits) while photographs are mostly made of three colors, each coded on 8 bits.
Consequently, there is much more information in AFM images than the simple chromographic transcription that is usually exported. Another consequence of the differences between these two kinds of images is that classical image processing techniques are not the best for AFM images. Even more importantly, the human eye is not a good judge of image quality after the transformation from 32 bits to 16 bits.
In this study, we have applied a classical image processing, the Laplacian operator, on two AFM datasets: a single protein (image below, a) original image, b) image plus the laplacian mask x1, c) image with a laplacian weight, d) image plus the laplacian mask x5) and the TMV.
Using a specific computation of "image visibility", it has been shown that the best treatment for both types of molecules are different. Consequently, despite the fact that the applied image processing functions are well known to the specialists, it remains very important to test different variations in image processing tools and it is also extremely important to report these tests in publications.
The Laplacian weight formalism was developed by Wendy Chen and the AFM images of proteins were obtained during the Cost Action TD1002 by Jean-Marie Teulon and Christian Godon.
Chen SWW, Teulon JM, Godon C and Pellequer JL (2016) Atomic Force Microscope, Molecular Imaging, and Analysis. J. Mol. Recognit. 29: 51–55.