Image Processing

Image processing As the speed, capability, and economic advantages of modern signal processing devices continue to increase, there is simultaneously an increase in efforts aimed at developing sophisticated, real-time automatic systems capable of emulating human abilities. Modern digital technology has made it possible to manipulate multidimensional signals with systems that range from simple digital circuits to advanced parallel computers. The goal of this manipulation can be divided into three categories: • • •

Image Processing image in -> image out Image Analysis image in -> measurements out Image Understanding image in -> high-level description out

Pattern Recognition PR is a science concerns the description or classification (recognition) of measurements, usually based on underlying model. The two major approaches to pattern recognition are the statistical (or decision theoretic) and the syntactic (or structural) approaches.

(Statistical)

Classification algorithm Observed world data

Sensor (camera)

Feature extraction algorithm

Classification

(Syntactic)

Description algorithm

Description

Image analysis We live in the three-dimensional world and energy is converted into two-dimensional entities called image by either our visual system or an electronic sensor, such as camera. Image analysis may be accomplished in a number of ways. For example, low-level features may be extracted from the grey-level image data, and this feature information processed sequentially, at increased high level. This is the bottom up or data directed approach. Conversely, we might hypothesize at the highest level scene characteristics and then proceed sequentially toward the low-levels, until the raw image grey-levels have been reached. This is an example of top-down approach. In practice, algorithmic approaches employ both of these approaches.

Physical world

Scene models Top - down

.

Scene instances

**Assumes viewing geometry and sensor model (e.g. ,p.p. Transform)

* Image instances Unification

** And other a priori information

** Input image (to be analyzed)

Bottom - up

Input image

Preprocessing operations (e.g. noise reduction)

Enhancement operators (e.g. edge detection)

Feature extraction (e.g. region segmen.)

Matching/ description (e.g. classification)

Image based results

Possible algorithmic interaction

Motion control

Captured image

Position feedback

Video Capture card

DSP (filters, Prog., etc..)

Raw im age Filtered Image

Host computer

One Image every 5 sec.

Motion control Image processing

Typical Control System with Image Processing Capability