> From: "Lloyd A Treinish" <[EMAIL PROTECTED]> > Date: Fri, 18 Jan 2002 12:35:33 -0500
> > > Well, let me ask some questions... > > 1. Are the images registered and navigated? That is, do you have > locations for the pixels in each image in some coordinate system? That's > not something that DX does out of the box, although you could add the > ability to do so. Yes. > > 2. If 1. is yes, is the coordinate system the same for both images? No. That is the reason I was asked to look at this problem. The coordinate systems are supposed to be the same but different microscopes from different vendors put 0,0 in slightly different places. > > 3. Are the images the same size? Yes except for the changes which are to be isolated. > > 4. If 1., 2. and 3. are yes, are the images coincident? No. They are supposed to be, but are not. Two images taken on one microscope and then immediately on another will have some error. It looks as if the angle from the observation point is different. This does cause both a shift in size and location which is minimized in the image center and maximized at the edges. > > Here are some things that you could so: > > If 3. or 4. are yes. You could do pixel-by-pixel operations with Compute > from trivial like differencing to complex comparisons. But they may only > make sense if 4. is yes. Otherwise the pixels would be of different sizes > in physical coordinates. > > If 1. and 2. are yes and 3 is no, then you can do overlays and > interpolation between them, the latter would enable you to do > pixel-by-pixel operations. If 2. is no, then you would have transform one > to the coordinate system of the other or both to a 3rd coordinate system > first. > > Independently of the above, you have the ability to image processing, > coloring, rendering, etc. in each of the images separately. > > These sorts of problems occur with remotely-sensed images in earth, > environmental and space sciences. Most texts on that subject will cover > topics of image processing, registration, classification and analysis. I sensed that. That is why I used the analogy. Thanks I will look for texts on that subject. > > > > Robert Lopez <[EMAIL PROTECTED]>@opendx.watson.ibm.com on > 01/17/2002 03:55:39 PM > > Please respond to [email protected] > > Sent by: [EMAIL PROTECTED] > > > To: [email protected] > cc: > Subject: [opendx-users] diff-ing similar images > > > > > I have been asked to find out what it would take to do > a certain project which is essentially extracting the > differences between two microscope images. I am way > over my head on this request and I will have to learn a > lot. I wonder if opendx can be used to do all or some > of it. > > An analogy of the microscope image problem might be > easier for me to communicate. > > Imagine flying an airplane over a city. At a point in > the air a photo is taken of a city below. Then a year > later flying over again and taking another photo. There > would be differences in the photo due to what changed > on the earth, but there would also be differences due > to other reasons such as weather, different airplane > positions, different camera settings etc. So there > would first have to be some manipulations just to get > the two images to overlay correctly. Then some lighting > artifacts from the weather have to be removed. Then > everything in the first image that is also in the > second image would have to be removed. The remaining > changes are all that should appear in the final image; > with perhaps some other coordinate information to > determine where the differences are in the city. > > I am also interested in references to the literatures > which cover ways to do this. Thanks. > > > [EMAIL PROTECTED] Assembly and Test Organization, Philips Semiconductors Mail Stop 20, 9201 Pan American Fwy. N.E., Albuquerque, New Mexico 87113 USA [P: (505)822-7112 F: x7237]
