Description
Part I
Consider the image shown in Fig. 1.
Fig. 1. “PandD.tif”
a) Utilizing the Hough transformation, design and implement an algorithm that detects (localizes) all coins in the
image shown in Fig. 1. Each detected coin must have a corresponding label mask in the output image to indicate
which pixels of the input image belong to it (à one, unique label per coin).
b) Design and implement a method to automatically classify each coin into pennies and dimes. Your algorithm
must produce a mapping of coin labels produced by a) to classes “penny” and “dime”. Use image-derived
features like color for classification. In your report, show and discuss your results. Explain the rationale behind
your approach.
Note that your grade will depend on the detection performance of your algorithm.
Part II
Generalize the algorithm developed in Part I as follows. The goal is to implement an algorithm which is able to detect
and classify penny, dime, and quarter coins in images that were acquired with varying distance between camera and
coins. For this purpose, perform the following steps.
a) Image acquisition – Acquire 6 new images that contain several coins. Make sure that samples of all three coins
are included in each image acquired. Furthermore, make sure that you have 3 images with “simple” and 3
images with more “challenging” background in your image set, each with a different scale (Fig. 2).
b) Algorithm development – Using the image dataset, expand the algorithm developed in Part I so that it can
successfully handle this multi-scale and multi-coin image analysis problem (detection and classification). In your
report, show and discuss your results. Explain the rationale behind your approach. Your grade will depend on
the detection performance of your approach.
Fig. 2. Examples for images with “challenging” background and different scales.
Deliverables:
Submit your assignment by using the Dropbox on ICON. The files to be submitted include the function(s)/script(s) you
have written and a report describing your work. In your report, discuss the rationale behind your approach as well as
any problems you encountered. Discuss the results you obtained. Also include a “conclusions” section, in which you
describe what you have learned from your experiments. For the paper, any common document file format (e.g., PDF) is
acceptable. Collect all your files and compress them using zip, tar, etc. and submit the compressed file. Note that one
submission/report per team is sufficient.
Furthermore, be prepared to present, demonstrate, and discuss your implementation in the last week of the DIP lecture.
More details will be provided in December.
