Intensity Transformation and
Spatial Filtering
19.08.2020
1:30 pm
Intensity Transformations Functions
• Image Negatives
• Log Transformations
• Power Law (Gamma) Transformations
• Piecewise-Linear Transformation Functions
• Contrast Stretching
• Intensity-level slicing
• Bit-plane slicing
Introduction
• r= value of pixel before processing
• s=value of pixel after processing
• s=T(r)
• Stored in one dimensional array
• Mapping from r to s - table lookups
• 8 bit – 256 entries
Image Negatives
• [0, L-1]
• s=L – 1 - r
• Negative – reversing the intensity
level
• Enhancing white or gray embedded
in dark region
Log Transformations
• s=clog( 1 + r)
• Narrow range of low intensity
• Inverse Log transformation
• Compresses the dynamic range of
images with large variations
• Pixel value in Large range – Fourier
Spectrum.
• Intensity detail may lost in Fourier
Spectrum
Power Law
• Gamma Transformation
• s=crγ
• Measurable output
• Input is 0 s=c(r+ε) γ
• Issue of display calibration
• Identity transformation – c = ϒ = 1
• Gamma Correction
Power Law
• Simply intensity map
• Gamma correction – straightforward
• S=r
1
2
.
5
= r0.4
• Input to the monitor
• Same intensity

intensitytransformationspatialfiltering.ppsx

  • 1.
    Intensity Transformation and SpatialFiltering 19.08.2020 1:30 pm
  • 2.
    Intensity Transformations Functions •Image Negatives • Log Transformations • Power Law (Gamma) Transformations • Piecewise-Linear Transformation Functions • Contrast Stretching • Intensity-level slicing • Bit-plane slicing
  • 3.
    Introduction • r= valueof pixel before processing • s=value of pixel after processing • s=T(r) • Stored in one dimensional array • Mapping from r to s - table lookups • 8 bit – 256 entries
  • 4.
    Image Negatives • [0,L-1] • s=L – 1 - r • Negative – reversing the intensity level • Enhancing white or gray embedded in dark region
  • 5.
    Log Transformations • s=clog(1 + r) • Narrow range of low intensity • Inverse Log transformation • Compresses the dynamic range of images with large variations • Pixel value in Large range – Fourier Spectrum. • Intensity detail may lost in Fourier Spectrum
  • 6.
    Power Law • GammaTransformation • s=crγ • Measurable output • Input is 0 s=c(r+ε) γ • Issue of display calibration • Identity transformation – c = ϒ = 1 • Gamma Correction
  • 7.
    Power Law • Simplyintensity map • Gamma correction – straightforward • S=r 1 2 . 5 = r0.4 • Input to the monitor • Same intensity