---

CIS 9601 Spring 2010

---

Homework 1: Image Retrieval using Edge Histograms

---

Please build a simple image retrieval system based on spatial edge pixel distribution. For an image I, create its edge image using the Canny Edge detector. Partition the image into m x n (e.g. 3x4) tiles. Count the edge pixels in each tile. This defines your edge histogram. Create the edge histogram for all pictures in the DATABASE. The database consists of 30 images (5 image classes, 6 images each). Use each of the pictures in the database as query, and show the top 5 results. The similarity value between the images is defined by the histogram similarity.

You have the freedom to experiment with different histogram distance measures and different patitioning refinements. Please report your experiences.

Enjoy, and good luck!