Thread View: sci.image.processing
3 messages
3 total messages
Started by Maureen Cain
Fri, 28 Aug 1992 22:39
Image Segmentation
Author: Maureen Cain
Date: Fri, 28 Aug 1992 22:39
Date: Fri, 28 Aug 1992 22:39
22 lines
1502 bytes
1502 bytes
I have an image that was taken through a microscope of a lead sample. It clearly (to the eye) has two distinct regions, one light and one dark. However, there is a lot of noise in both regions. I want to extract the contours from the image defining the light and dark regions. However, because of the noise, standard edge detection algorithms (Canny, Sobel, etc) do not work very well, even after the image has been preprocessed with a median filter. I need to get a closed contour from the image as if I took a pencil and traced it off the image. Is there a way to segment the image using contrast and variance stats rather than using and edge detector? Or a combination of both? Any help would be appreciated. Thanks Maureen Cain ttcain@vangogh.lerc.nasa.gov (216) 433-8928 -- ----------------------------------------------------------------------------- Maureen Cain phone: 216-433-4000 NASA Lewis Research Center Cleveland, Ohio 44135 email: ttcain@vangogh.lerc.nasa.gov
Re: Image Segmentation
Author: John C. Schultz
Date: Fri, 28 Aug 1992 18:15
Date: Fri, 28 Aug 1992 18:15
27 lines
1910 bytes
1910 bytes
In article <1992Aug28.223922.11107@eagle.lerc.nasa.gov> ttcain@vangogh.lerc.nasa.gov (Maureen Cain) writes: I have an image that was taken through a microscope of a lead sample. ... median filter. I need to get a closed contour from the image as if I took a pencil and traced it off the image. Is there a way to segment the image using contrast and variance stats rather than using and edge detector? Or a combination of both? Any help would be appreciated. Thanks The traditional way, which even works sometimes :-), is to use difference of gaussian (or DOG) filters. The basic premise is to size two lowpass operations such that subtracting the two filtered images results in a bandpass image retaining the object with the desired spatial dimensions. The absolute and relative size of the bandpasses is therefore important. A reasonably good approximation is to take the difference of two box-car filters (rectangles). The math is not as nice as the gaussian filters but the programming is fairly trivial. After thresholding, morphological operations can then clean up the resulting image. -- John C. Schultz +1 (612) 733-4047 schultz@halley.serc.3m.com How to include the taste of Glendronach in a multi-media system?
Re: Image Segmentation
Author: Brad Cote
Date: Sat, 29 Aug 1992 00:32
Date: Sat, 29 Aug 1992 00:32
18 lines
912 bytes
912 bytes
ttcain@vangogh.lerc.nasa.gov (Maureen Cain) writes: > Is there a way to segment the image using contrast and variance >stats rather than using and edge detector? Or a combination of both? Any >help would be appreciated. Thanks >Maureen Cain This might be helpfull: "Image Segmentation by Succesive Background Extraction," by A. Rodriguez and O. Mitchell, Pattern Recognition V24 #5 pp 409-420, 1991. Brad Cote' University of California, San Diego STARE Project
Thread Navigation
This is a paginated view of messages in the thread with full content displayed inline.
Messages are displayed in chronological order, with the original post highlighted in green.
Use pagination controls to navigate through all messages in large threads.
Back to All Threads