Assignment 1 - Image Preprocessing & Segmentation
DUE 19.11
To all relevant questions, provide the block diagram of the algorithms
and the code
1. Histograms (30%)
1. Open Head1.bmp
image, and present it. Show the histogram of the image.
2. Perform contrast enhancement on the image in two ways:
-
Manual stretching (scaling using a linear transformation: select a min
& max range of the image and scale it to 0-255 range).
-
Histogram equalization via Cumulative Distance Function (CDF). Write your
own code.
-
Discuss the two output images and histograms.
3. Assume that the gray-levels r have the probability density
function:
-
Use the CDF transformation to find S=T(r).
-
Show (mathematically) that the resulting density Ps(S) is uniform.
2. Edge detection (30%)
Open
Coronary.bmp image and
present it. Use different filters to perform edge detection on the image
and extract the vessels' boundaries. Filters
should include the Sobel operator and Laplacian-of-Gaussian
(LOG).
-
Show resultant edge images and discuss the results.
-
What are the difficulties in extracting a closed heart boundary, the vessel
boundaries, and the stenosis size (diameter)? List.
-
Suggest additional algorithms you may use to achieve these goals.
3. Region Growing and Region Splitting
(40%)
3.1 Region Growing
Open
Isch_head.bmp
and present it. In the image you see a ``white” region. This region is
an ischemic region inside the brain that needs to be
detected and measured.
-
Use histogram thresholding to separate out the region of interest. Show
resultant binary image and discuss what threshold you chose.
-
Code a Region Growing algorithm and use it to measure the area of the ischemia
(in pixels). Select a seed point inside the region.
Show resultant region following the Region Growing algorithm.
3.2 Region Splitting
-
Segment the image shown below by using the split-and-merge procedure. Let
P(Ri)=TRUE
if all pixels in Ri have the same intensity.
-
Show the quadtree corresponding to your segmentation.
