Image Processing
Digital Images
Discrete representation of spatial and intensity values in a 2D matrix.
Raw images are often:
- Extremely noisy
- Contain invisible bands of the EM spectrum.
- Contain too much information (that needs to be cropped out)
Image Sensing & Optics
Images are a signal so it is collected using a sensor.
- Light is reflected off, or emitted by, objects.
- Photos pass through the optics.
- Photos are focussed onto the focal plane.
- Electronically or chemically.
- The energy of photos is importance when forming images.
Not all images are in the visible spectrum so we can use other wavelengths, ultrasonic and so on to create images.
Attenuation
IR Cameras
They are good when there isn’t a large amount of ambient light.
There are three different types of IR:
- Near IR 0.7um - 2um
- Little thermal emission, but can use laser illumination.
- Mid-wave Infrared 3um - 5um
- Good for hot objects as there is less background thermal emission.
- Long-Wave Infrared 8um - 14um
- Good for room temperature (thermal LWIR emission peaks at 300K)
Low Light
Visible band cameras only work well when there is ambient light.
This means they are not good at night.
To correct for this we can use an amplifier:
- This introduces a lot of noise.
- Tends to make the camera sensitive to glare.
Tools for Digital Image Processing
- MATLAB (licensed)
- Through image processing toolbox.
- ImageJ (free)
- OpenCV (free)
Typical Image Processing System
graph LR
pp[Pre-processing] --> mp[Main Processing]
mp --> pp2[Post-Processing]
Pre-Processing:
- Basic Correction
- Basic Noise Removal
- Background Removal
- Alignment
- Format Changing
- Transformation
Main Processing:
- Advanced Correction
- Advanced Noise removal
- Segmentation
- Detection
- Searching
- Feature Extraction
- Recognition
Post Processing:
- Image Formation
- Format Changing
- Result Representation
- Report Generation
You will gain significantly better results if you pre-process.