Description
ELEC4630/7463 Image Processing and Computer Vision Assignment 1
1. Automatically locate the number plate in the following image. (Available in
numberplates.zip from Blackboard). You may try a 2D cross-correlation with a
template matching the plate border, but you will probably need to ‘chamfer’ the
template by convolving with a Gaussian or some other blurring function. You’ll
also need a good edge detector such as the Canny Edge detector. Test your method
on some other example car images from numberplates.zip.
In your report discuss methods used, problems encountered, performance, and
possible solutions. Comment on the problems encountered in plate extraction and
the difficulties in designing a general plate extractor.
(10 Marks)
2. You are provided with a video of a pantograph which provides the electrical power
to a train (Eric_Video.avi). The electrical cable slides across the carbon brush of
the pantograph in a zig zag manner. Your task is to graph the position of the power
cable on the pantograph over time from the video. Note that both the suspension
cable and the power cable are visible and that we only want to track the power
cable. Note further that at some times in the video there are two or more power
cables visible. We only want to track the lowest cable which is powering the train.
See if you can work out a way to separate the cables.
(10 Marks)
Remember to demonstrate your solution to the tutors in class so that they can assign
you a mark. The aim is not to get a perfect answer, but to show that you know how
to approach the problem and can describe the image analysis challenges. Therefore
your assignment needs considerable focus on method selection, algorithms, and
discussion of the results obtained. The assignment should be presented in the form
of a brief report including code, images, and descriptive text.
(Total: 20 Marks)
ELEC4630/7463 Image Processing and Computer Vision Assignment 2
1. Use the Hough transform to find the four straight lines of the water feature border in
this image and project them back onto the original image. Note that the minarets
are tilted out by a few degrees, so they look vertical to the observer and to avoid
them crashing into the dome during an earthquake. See if you can estimate the
angle of tilt of each tower.
It is also possible to code this using the Radon transform in Matlab, see the tutor.
(Image available on Blackboard site). Note there is a Hough Transform function in
Matlab, but I would like students also try to implement this transform from first
principles and then compare the results.
(5 Marks)
Figure 1 Taj Mahal
2. Design an algorithm to determine the cross-sectional area of the heart in this MRI
Image Sequence of a working heart. An example image is given; the other images
are on the Blackboard site along with sample segmentations. By examining the
cross-sectional area over time, a cardiologist can determine the efficiency of the
heart. This is especially important diagnostic information for cardiologists after
patients have had heart attacks.
Suggest you examine snakes, morphology and related methods to solve this
problem.
See site for book on related work in segmentation.
Sambhunath Biswas and Brian C. Lovell (2007) Bezier and Splines in Image
Processing and Machine Vision, Springer
For your report, you should determine area of the heart in each image example, and
provide commented code, a description of your methods, and why you chose them.
Your algorithm should be able to work on the entire set of images without
modification or parameter adjustment. If possible, produce a movie of your
segmented image stream and a graph of cross-sectional area versus time.
Figure 2 MRI Image of Heart
(15 Marks)
(Total 20 Marks)
Outer wall of left ventricle
Inner wall of left ventricle
ELEC4630/7463 Image Processing and Computer Vision Assignment 3
1. Segment some sample ANA IFF HEp 2 cell images using appropriate methods of
your choice (thresholding, morphology, etc). The example images on Blackboard
represent borderline and high positive samples. Comment briefly on the challenges
and the results you obtained. Design a suitable metric to measure how well you can
segment these cells. Explain your method.
(10 Marks)
Figure 1 ANA IIF HEp-2 Cell Images: Positive Example
2. Design an algorithm to determine the length of a piece of string from images. The
string can form loops and cross over itself. You are provided with sample images of
Strings 1 and 2 taken under the same conditions and you will be given the physical
length of the string for calibration. For your assignment, you should determine the
length of the string in each image example, and provide commented code and a
description of your methods. Your algorithm should be able to work on the entire
database including string 3 without modification or parameter adjustment.
(7 marks)
String 1 = 13cm String 2 = 15.5cm String 3 = ??cm
3. Using your algorithm from Question 2 or otherwise solve the following problem.
Try to develop a GUI so that clicking on a string will show the complete path. The
scanned image is available on Blackboard.
(Total 3 Marks)
(Total 20 Marks)
ELEC4630/7463 Image Processing and Computer Vision Assignment 4
1. Stereo reconstruction is only useful for reconstructing a simple scene from two nearby
views. It can be difficult to extend to complicated scenes with occlusion and to a large
number of cameras or views. Volumetric reconstruction is one way to handle a large
number of cameras with arbitrary viewpoints. In this assignment you will be required
to fuse the data from 36 viewpoints of an object into a detailed volumetric model of
that object. An example result is given below.
The inputs:
36 viewpoints of a toy dinosaur spinning on a turntable (10° per image).
36 projection matrices which determine how each spatial point maps onto the image
plane.
The output:
A volumetric model of the dinosaur
(To be submitted) 4 new viewpoints of the reconstructed dinosaur which do not
coincide with any of the original 36 viewpoints. Try to texture map the images.
The method you are required to implement is known as “shape from silhouette”. This has
been covered in lectures. Images and some hints are available from Blackboard as dino.zip.
(10 Marks)
2. Build a face recognition system based on the eigenface technique to recognize faces
from the database on the website. You should be able to obtain about 96% correct
recognition. Supply the code listing, explain your method, and performance analysis.
Higher marks can be obtained by building a nice gui interface in Matlab. Face images are
available from Blackboard. Details of the Eigenface method are in the book chapter “Face
Recognition for Data Mining” on Blackboard.
(10 Marks)
(Total 20 Marks)
