You are on page 1of 2

Image Project Overview: for this project you will use multiple image files, provided by the instructor,

and perform segmentation, representation and recognition processing techniques to analyze and classify each of the images. The following provides guidance for completing and submitting the results of your analysis. Details: a. Unzip the file provided with the image project assignment and select at least three different classes you want to identify. For example, the zip file contains 650 images of the letters from A-Z. In most cases, you will see 25 different variants of each character. You should select at least 3 letters (e.g., X, R and D) for your classes. Then you should select at least 5 representatives from each of those classes. These will be the images you will use for analysis and building your classifier. Note: the images may have the 1s and 0s reversed so you should verify and swap these values if needed using this Octave command: A1=imcomplement(A1); b. Using Octave, plot each of the images you have selected (using imshow()) and copy these images to a Word document. Be sure to label each image showing its original filename (e.g., a1. bmp). c. For each image, using Octave, perform the following calculations: 1. 2. 3. 4. 5. Boundary (using bwboundaries()) Freemans Chain code (using fchcode()) MPP Boundary length Boundary Diameter

d. Plot the MPP for each image and describe the similarities seen in the same image types. Show the MPP and associated similarity discussions in your word document. e. Using Octave, design a minimum distance classifier, with a pattern vector of your design that best differentiates the classes you have selected. You will need to perform many iterations of design to optimize your pattern vector. For example, you will probably start with a vector of only 2 elements and then add other elements as needed to help best identify your classes. Consider variations on a theme such as using Freeman Chain codes or the difference in freeman chain codes and other variations that interest you. f. After you have a classifier that optimizes the classification of each of the images, document the results in a word document, showing the decision functions for each of your classes. You should also describe in detail how you determined the optimum classifier and what different models you used. Also, provide the exact classification results in a table like the one shown below. Notice the left column shows the total count for each class. Then each of the other columns list the count of classes

that were classified into G, K or T (Note: your columns will look different depending upon the classes you selected and the number of images you selected). :

Class G (Count=5) K (Count=5) T (Count=5) Summary

G K T 4 (80%) 1 (20%) 0 (0%) 1 (20%) 3 (60%) 1 (20%) 0 (0%) 0 (0%) 5 (100%) A total of 12 out of 15 images were correctly classified (80%)

A perfect classifier would look like this (However, you arent graded if your classifier is perfect, but how you work to optimize yours. In fact, I would suspect most classifiers will misclassify some images): Class G (Count=5) K (Count=5) T (Count=5) Summary G K T 5 (100%) 0 (0%) 0 (0%) 0 (0%) 5 (100%) 0 (0%) 0 (0%) 0 (0%) 5 (100%) A total of 15 out of 15 images were correctly classified (100%)

Deliverables: You should include one Octave m file that solves all of the above parts for this homework. You should name it yournameimage.m. Be sure your m file works without issue and accurately provides the outputs and plots as needed for this assignment. In a word document, describe the results and your analysis and observations and other data required in the description above. Your word document should be named yournameimage.doc (or yournameimage.docx).

You might also like