Archived posting to the Leica Users Group, 2001/07/24[Author Prev] [Author Next] [Thread Prev] [Thread Next] [Author Index] [Topic Index] [Home] [Search]
> Austin writes: > > > If you want to detect a line that has a > > width of 0.009mm, which is the width of the > > line in a 53 lp/mm set, which is 106 lines, > > you need to sample at slightly more than > > 2x that, or < .0045 mm/sensor in order to > > detect the line RELIABLY, ie, every time. > > Exactly 2x is fine. Er, yes, I know...that's why I said it. Which means you need a scanner of 5400 SPI to reliably detect your 53 lp/mm... > > Your are flawed in thinking that you can > > detect a .009mm line with a .009mm wide sensor > > RELIABLY. > > You haven't defined what you mean by "reliably." Er, go to my paragraph above, where I wrote "RELIABLY, ie, every time". It's a standard term. "Yielding the same results in statistical trials." In other words, always returns the same result. > > Draw a picture for your self, and you will > > see that if you make your line pairs which are > > 0.009mm per line, straddle the .009mm sensors, > > such that 1/2 of a line is over one sensor, > > and 1/2 over the other, you may not detect any > > lines at all, you will get gray. > > The gray _is_ the detection of the line; it is gray because the > line is only > half present in each pixel. This is exactly the expected result, and it > corresponds to true detection of the lines. If the entire image is gray, that, obviously, is not resolving any lines, since there is no contrast...and it requires contrast for discrimination. We have been talking about RESOLVING the line, not just detecting it. > > Please think about this before replying. I've > > spend the past 20+ years designing digital > > imaging systems...so this is not just an academic > > exercise for me. > > I've spent the same amount of time working with information > theory, and I know > exactly whereof I speak. How does that apply to this? I don't believe it really does, because you've been arguing with me about something that is just known, and unquestioned, by people who design digital acquisition systems. You may be a wiz in "information theory", but I'd suggest that you don't design any scanners until you get a real understanding of the mechanisms involved...or do design a scanner and learn how it's done! No better teacher than experience. > Keep in mind that a practical > understanding and a > theoretical understanding are not necessarily coincident, nor is > one necessarily > a subset or superset of the other. I have clearly shown, and I'd have to believe, given your response here, that you agree, that you need to sample at >2x to "reliably" RESOLVE a line. This is both theory and practical understanding, they are the same in this case. There is a third, and that is experience. You obviously understand some of this stuff, but you are missing some understanding too. Nothing wrong with that, it appears this is not your field of primary expertise.