If the information is there you can get it out. It may be poor in quality. And one wouldn't use the kind of image processing I mention on bacterial cells. Perhaps on some of their ultrastructural components. The calls are not ordered enough and they are far far too big. The purple membrane of photosynthetic bacteria has been studied by electron microscopy and image analysis. And who wouldn't like perfect images and clean spectra?
Don Don Williams ___________ Dr E D F Williams http://personal.inet.fi/cool/don.williams Author's Web Site and Photo Gallery Updated: March 30, 2002 ----- Original Message ----- From: "Gregory L. Hansen" <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Wednesday, January 08, 2003 9:31 PM Subject: Re: pentax-discuss-d Digest V03 #8 > Dr E D F Williams said: > > > It does not matter how noisy a signal is, if the information is there it can > > be retrieved. In Electron Microscopy the images are often terribly noisy. > > For ordered structures Fourier transforms, rotational symmetry, or a > > combination of methods is useful. I have programs to do things like that and > > if I can find a decent electron micrograph of a virus I'll try to prepare > > some images that illustrate the cleaning of an image. Image processing can > > be done in real time on an optical bench, but its more difficult. > > It matters, but it depends on the application. Pictures of virii are > usually shown in black and white, for instance, and bacteria stained to > present false colors. What's important there is to see structure. In my > case, I've been trying to extract signal that's a factor of 100 smaller > than the noise, and it's not easy and I don't really trust the results I'm > getting. There's nothing to do but keep taking data until the statistics > favor me. In terms of photography, if you have too much noise you might > be able to clean up a picture and clearly see the features of a dog, but > lose much of the texture of fur and other small details. It's always > better when your raw data is as clean as possible. >
