Digital Image Processing Part 1 Introduction
Jan 17, 2016
Digital Image Processing
Part 1
Introduction
The eye
The eye
• Retina is covered by photoreceptors: cones and rods • Fovea contains cones only (most vision here)• Outside fovea, the retina is mostly covered by rods
• Cones: primarily color perception• Long - more sensitive to long wavelength (red)• Medium - more sensitive to medium wavelength (green)• Short - more sensitive to short wavelength (blue)
• Rods: highly sensitive to intensity and black/white vision, responsible for vision under dark-dim condition
Receptor distribution
• High density cones around 0 (1.5 mm)• Some rods in fovea but most outside• About 6 – 7 million cones, 75 – 150 million rods
Test Blind Spot
• Cover one eye
• Focus on the X and move closer to the screen
• The spot eventually becomes invisible but moving closer still enables you to see it again
Visible light
Colour perception
There is no colour in light but the stimulation of the rods and cones by specific frequencies creates a visual representation of colour in the brain
Red receptors cover a significant portion of the green band. Green is perceived more strongly than red and both more than blue
Perception is around: 0.59G, 0.3R and 0.11B
Colour blindness
• Defective X chromosome
• Men have XY women have XX so higher probability of having one good X
• 8% of males are red / green colour blind
• Women may be able to detect subtle red green differences better through combination of XX
Colour blindness test 1
Colour blindness test 2
Colour blindness test 3
Camera and Eye comparison
Image Sensor Operation
• Charge Coupled Device image sensor
• Background is image sensor zoomed in
• CCD measures brightness• Tiny lenses direct light onto filtered
photosensitive regions• More green than red to better match the eye
Bayer filter
• Eye responds mostly to green so as many green filters as red+blue
• Demosaic – interpolate a single pixel colour by interpolating nearest neighbours as each pixel only records one colour so the actual colour at that point is the average if it and the surrounding pixels
CCD Operation
CCD Operation
• Charge is moved down 1 row at a time then clocked out to an amp and A to D converter
Colour models
• Additive – light, e.g. Computer Monitor– Primary: Red, Green, Blue– Secondary: R+G=Y, B+R=M, B+G=C– R+G+B = White
• Subtractive – pigment e.g. printer– Primary: Yellow, Magenta, Cyan– Secondary: R,G,B– Y+M+C = Black
Colour Properties
• Hue (color of the light - dominant frequency)
• Brightness – perceived intensity of the light
• Saturation (purity) – a measure of the degree to which a pure color is diluted by white light
• Chromaticity refers to Hue and Saturation
Monitor & Printer Colour Models
RGB colour cube
HIS Model (human model)
HSI Conversion
• Hue, Saturation and Intensity are separated
• Several operations work on intensity only so this model is ideal for them (luminance or grey level). For example, brightness and contrast modification
• Converting RGB to I is trivial: I=(R+B+G)/3
• H and S are not trivial
Grey Scale
Sampling
Sampled image
Brightness and Contrast
• Grey scale and histogram of pixel values
Dark and Light Images
Low Contrast
To lighten or darken, shift the distribution left or right
To increase contrast, stretch the distribution over a wider range
Good Contrast
• Almost full dynamic range used. Could contrast stretch slightly
Point Processing - Brightness
• Map current pixel values into new values. Useful for brightness, contrast, posterise and threshold.
Brightness adjustment• To increase brightness, add or subtract
constant from each pixel colour value
• Need to clamp at MAX (255) and MIN (0)
Convert to greyscale• Produce mean brightness levels:• Multiply each pixel colour value by:
0.59xGreen, 0.11xBlue and 0.3xRed• Need to clamp at MAX (255) and MIN (0)
Point process contrast