Assignments

Assignment 4, Fall 2018

Posted here

Fall 2016

1-A report on the hyperspectral cameras and their applications. Your report should include technical details as much as possible. Deadline October 12th, 2016- 12:00 PM.

2-—One of the important methods for corner detection is Harris Corner Detector. —Write a report describing the details of the method step by step —Implement the method using MATLAB. Explain the result of each step. Justify your results if they are not the expected values.

Deadline October 26th, 2016- 12:00 PM.
 

3- Hit-or-miss morphological operation can detect clearly defined shapes in images. However, in real-life applications the images suffer from noise or other effects which reduce the quality of the image and makes the boundaries of the objects unclear. Besides, the size and orientation of the objects cannot be predicted in advance. 

Modify the original hit-or-miss algorithm to make it less sensitive to noise, and scale and orientation invariant. Apply your modified method on a sample binary image and discuss the results.

Deadline: November 23, 2016

4- In the paper given here the authors claim that they have improved the Fourier description for object representation. Prepare a report answering the following questions:

a) What are the main differences between the original Fourier descriptors and the improved ones.

b) How does the improved method solve scale problem? Use your own example to describe their solution.

c) How the rotation problem is handeled?

d) Is the proposed method sensitive to noise? Explain.

Deadline:  December 7th, 2016

Fall 2015

1-Prepare a report on how a hyper-spectral camera works. Give main advantages and disadvantages, and mention its probable applications

2-Apply Hough transform to an image and draw the detected lines on the original image

3-To segment an image into multiple regions, the texture properties of each region can be used as a feature. In pixel domain gray level co-occurrence matrix can be defined for sub-images and the statistical moments can be used for segmentation.

Assume a sample image (gray level) with multiple regions is given. Define a window of nxn (choose appropriate value for n) and compute texture moments for the pixels faling under the window. For segmentation, put window on each area and find the moments. This will provide the expected value of the moments for each region. Then find the moment values for some region and sub-image it to one of the regions having the closest moment value.

4- Assume an imaging system is used to inspect the accuracy of washers produced in a manufacturing plant. Assume the image of an error free washer is provided. Develop an algorithm using morphological operators for this problem. You may download test images from here and here.