Unformatted text preview: neighbor classifier. You can use 2 images for training samples and 1 image for test. i. Compute center of gravity for each signature image. Could you classify different classes using this feature without any misclassification? ii. Compute width and height of signatures. First, classify different classes based on only width of signatures. Then do classification only based on height of signatures. Compute accuracy of classification for each feature. iii. Which feature is the most important? Why? iv. Now, classify signature classes using 3 features (center of gravity, width and height of signature)? How much is accuracy of this approach? v. Could you propose some other features for above problem? Apply your proposed features and compute accuracy of classification. . Hints: You can use some beneficial MATLAB functions: • graythresh • im2bw • bwlabel • bwboundaries • bwperim...
View Full Document
- Spring '10
- Image processing, Digital image, high quality grains