Some Elements of Image Analysis and Computer Vision

Some Elements of Image Analysis and Computer Vision Problem: extract “important features” from images Examples: • distinguish and categorize parts in an assembly line; • Optical Character Recognition (OCR) to sort mail, read labels, automatic filing • detect road shoulder lines, cars, obstacles in unmanned road vehicle • categorize human activities such as walking through, standing, depositing packages … from surveillance videos Challenges: develop appropriate mathematical tools.

Given an Image or a Video … … 3 main steps to Image Understanding: 1. Feature Extraction: 2. From Intensity and Color Information, extract relevant features such as texture, edges, motion, lines, circles … 2. Segmentation: Determine regions with similar features 3. Classification: Classify each region and extract relevant information type “A” cells type “B” cells

Image Segmentation 1 0 Segmentation Gray Level image Most often done by thresholding: A few Level image (most often Binary)

The first “raw” result is usually “noisy” We would like a solid “blob” like this: … but we get this: “broken” blob pixels which don’t belong to the region We need to do some “filtering”. Since we deal with binary data, we need to define binary set operations. This leads to Morphological Processing

Morphological Processing Define mathematical operations similar to filtering, for binary data. Typical results: MP

Why is this important: From this data the computer can detect an object and compute some of its parameters: detection

Basic Definitions: set of all real numbers x set of all integers n set of all pairs of real numbers set of all pairs of integers Relations between sets: inclusion also written as

Operations between sets: Union: Intersection:

Complement: Shift:

Morphological Operations Given a binary Image X and a mask B define: Dilation: new points old points

Erosion: