I was going through a paper on image segmentation, and a number of questions came to mind:
1. what does it mean to compute the gradient magnitude of a segmented image or an image?
2. Of what use is gradient magnitude in image processing?
3. What is the relationship between marker-controlled watershed segmentation and gradient magnitude?
1 Gradient computation of a segmented image makes no sense. A segmented image has two pixel values: foreground (the region of interest) and background. One can do it, and the result is the contours of the region of interest, but there are faster ways to get the contours. Gradient computation of a (non-segmented) image is a kind of edge detection; it enhances the areas in an image where the rate of change is high. Usually done by subtracting each pixel in a row from the previous pixel (i.e. horizontal gradient) and each pixel in a column from the previous one (i.e. vertical gradient). Applying Pythagoras to the horizontal and vertical images results in the gradient .
2 It is mainly used for edge detection: a pretty fast local operation.
3 I don't know. Watershed is independent of gradient calculation. Watershed can be done on the original image, and also on a gradient image. I don't know what marker controlled means in this context. Maybe watershed seed(s), i.e. the starting point(s) of a watershed operation?
Yes, thanks for your response Dr. Lambert Zijp. While I was reviewing the paper titled "MEDICAL IMAGE SEGMENTATION USING FUZZY C-MEANS CLUSTERING AND MARKER CONTROLLED WATERSHED ALGORITHM"( www.ijmer.com/papers/vol2_issue1/BN021408411.pdf ) I came across a method that initially segment an image using FCM and then apply Sobel Filter on the segmented image before applying marker-controlled watershed segmentation and this has been confusing to me.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing