Hi, I want to process a large number of greyscale images similar to the example below - which may contain variable amounts of noise. My objective is to identify the upper reflective band - partially highlighted in Red in those images - in order to extract information on its average width (absolute value is not significant, but being able to compare thicknesses of this layer across images is the critical task).
This "top band" consistently traverses the image left-to-right (in this displayed orientation) - but can be slanted or shifted; relying on band continuity can be important in images where the signal gets noisy (as in the 3rd image).
From your experience, what processing steps (/OpenCV algos) would you use to segment this top signal band/extract its thickness information ?
Finally here's also an example processed with a prototype algo that fails to account for continuity (bounds of the band are marked with white dots); I'd know to modify/fix it, but rather than adding a layer of filtering to our current approach, I'd like to tap the OpenCV community for new ideas on the best approach:
Thanks in advance for your ideas and suggestions !
Ivan