Edge Detection and Segmentation of Images
77
Quad-tree algorithms are often too computationally time consuming and less
accurate than the previously discussed method such as seed growing algorithm. This
is primarily due to the fact that they do not require seed points. However, in biomedical image processing, it is often the case that physicians have reliable estimates of
the seed points that can be used for simple seed growing methods. This fact rather
eliminates the need for quad-trees in many biomedical applications. In addition,
since in medical diagnostics physicians prefer to supervise and control the segmentation and classification steps (as opposed to completely relying on machine decision),
supervised region growing with seeds has proved to be more applicable than fully
automated quad-trees for medical applications.
4.4 SUMMARY
In this chapter, we discussed the computational techniques used for edge detection
and segmentation of images. While many edge detection and segmentation methods
are based on differences among pixels and regions, there are a number of methods
that utilize the similarities for segmentations. In this chapter, we covered the main
segmentation methods in each of the two groups. We also gave MATLAB examples
and simulations for the introduced methods.
PROBLEMS
4.1 L oad the image in the file “p_4_1.mat” and show the image. This image is a
fluoroscopic image of the heart and a catheter that is inserted in the blood vessels. In such images, it is desirable to extract the edges and lines representing
important objects such as blood vessels and catheter.
a. A pply horizontal Sobel mask to extract horizontal edges in the image. Show
the resulting image.
b. A pply vertical Sobel mask to extract vertical edges in the image. Show the
resulting image.
c. A pply Laplacian of Gaussian method to extract all edges in the image. Show
the resulting image.
d. A pply Canny edge detection method to extract all edges in the image. Show
the resulting image.
e. Compare the results of the previous sections and comments on the differences. Which method can highlight the catheter more effectively?
4.2 L oad the image in the file “p_4_2.mat” and show the image. This image is also
a fluoroscopic image of the heart and a catheter inserted in the blood vessels. As
mentioned earlier, in fluoroscopic images, the objective is to highlight the lines
formed by important objects such as blood vessels and catheter.
a. A pply horizontal line detection mask to highlight the horizontal linear
objects and show the resulting image.
b. Apply vertical line detection mask to highlight the vertical linear objects
and show the resulting image.
c. Apply the mask for detection of rising lines with the angle of 45° line to
highlight these linear objects and show the resulting image.
77
Quad-tree algorithms are often too computationally time consuming and less
accurate than the previously discussed method such as seed growing algorithm. This
is primarily due to the fact that they do not require seed points. However, in biomedical image processing, it is often the case that physicians have reliable estimates of
the seed points that can be used for simple seed growing methods. This fact rather
eliminates the need for quad-trees in many biomedical applications. In addition,
since in medical diagnostics physicians prefer to supervise and control the segmentation and classification steps (as opposed to completely relying on machine decision),
supervised region growing with seeds has proved to be more applicable than fully
automated quad-trees for medical applications.
4.4 SUMMARY
In this chapter, we discussed the computational techniques used for edge detection
and segmentation of images. While many edge detection and segmentation methods
are based on differences among pixels and regions, there are a number of methods
that utilize the similarities for segmentations. In this chapter, we covered the main
segmentation methods in each of the two groups. We also gave MATLAB examples
and simulations for the introduced methods.
PROBLEMS
4.1 L oad the image in the file “p_4_1.mat” and show the image. This image is a
fluoroscopic image of the heart and a catheter that is inserted in the blood vessels. In such images, it is desirable to extract the edges and lines representing
important objects such as blood vessels and catheter.
a. A pply horizontal Sobel mask to extract horizontal edges in the image. Show
the resulting image.
b. A pply vertical Sobel mask to extract vertical edges in the image. Show the
resulting image.
c. A pply Laplacian of Gaussian method to extract all edges in the image. Show
the resulting image.
d. A pply Canny edge detection method to extract all edges in the image. Show
the resulting image.
e. Compare the results of the previous sections and comments on the differences. Which method can highlight the catheter more effectively?
4.2 L oad the image in the file “p_4_2.mat” and show the image. This image is also
a fluoroscopic image of the heart and a catheter inserted in the blood vessels. As
mentioned earlier, in fluoroscopic images, the objective is to highlight the lines
formed by important objects such as blood vessels and catheter.
a. A pply horizontal line detection mask to highlight the horizontal linear
objects and show the resulting image.
b. Apply vertical line detection mask to highlight the vertical linear objects
and show the resulting image.
c. Apply the mask for detection of rising lines with the angle of 45° line to
highlight these linear objects and show the resulting image.