Contrast Enhancement

1
Imaging and Special Effects CE00376-2
Fac. of Comp., Eng. Tech. Staffordshire
University
Contrast Enhancement
Dr. Claude C. Chibelushi
2
Outline

• Introduction
• Histogram
• Histogram stretch
• Histogram equalisation
• Summary

3
Introduction
or
Contrast enhancement
4
Introduction

• Contrast enhancement
• changes amplitude of pixels to cover wider range
• possible techniques
• histogram stretch
• histogram equalisation
• spatial coverage of information used for
  transformation
• global based on whole image
• local based on part of image

5
Histogram

• Typically contains counts of pixels at each pixel
  amplitude (e.g. grey level)
• can provide local / global information about
  image
• hence sometimes used for image transformation
• e.g. contrast enhancement
• Many types of histograms
• other image attributes can also be counted
• e.g. edge orientation

6
Histogram

• 1D array
• size equal to the number of grey levels

Grey level
int histogramIM_DEPTH
7
Histogram

• Implementation
• for (row 0 row lt IM_HEIGHT row)
• for (col 0 col lt IM_WIDTH col)
• pixVal imagerowcol
• histogrampixVal

Requires initialisation to 0
8
Histogram

• Alternative implementation
• for (row 0 row lt IM_HEIGHT row)
• for (col 0 col lt IM_WIDTH col)
• histogram imagerowcol

9
Histogram Stretch

• Grey-level scaling (revisited)
• Pixel grey-level scaling
• changes amplitude of pixels to cover wider range
• Problems scaling might introduce
• amplitude overflow
• side effect possible grey-level shift
• Solutions first counter possible shift, then
  scale grey levels to fit within valid amplitude
  range

10
Histogram Stretch

• Grey-level scaling (revisited)
• Naïve contrast enhancement
• for (row 0 row lt IM_HEIGHT row)
• for (col 0 col lt IM_WIDTH col)
• if (imagerowcol lt 255 / SCALE)
• imagerowcol SCALE
• else
• imagerowcol 255

Inefficient
assumed to be positive
11
Histogram Stretch

• Histogram stretch contrast enhancement technique

12
Histogram Stretch

• Two transformations for simple linear stretch
• pixel amplitude shift
• to align current lowest pixel amplitude with
  desired lowest amplitude
• pixel amplitude scaling
• to stretch current histogram pixel-amplitude span
  to desired span

13
Histogram Stretch

• General linear mapping function
• for stretching grey levels (old value g) from
  range minold, maxold to minnew, maxnew

1. Grey-level shift
2. Grey-level scaling
3. Grey-level shift

• Exercise
• use above formula to modify code for naïve
  grey-level scaling
• no need to compute histogram

14
Histogram Equalisation

• Contrast enhancement technique
• Tries to obtain flatter histogram

15
Histogram Equalisation

• Pixel-amplitude reassignment requires mapping
  function
• possible mapping function

16
Histogram Equalisation

• where (if grey levels are to be stretched over
  full range)
• gnew new pixel grey level (old value g)
• t(g) count of pixels with grey level from 0 to g
• G number of possible grey levels (normally 256)
• N number of pixels in image

17
Histogram Equalisation

• Example - histogram prior to equalisation

18
Histogram Equalisation

• Example - calculations

19
Histogram Equalisation

• Example - histogram after equalisation

20
Histogram Equalisation

• Implementation histogram equalisation
• prevCumCount 0
• / Calculate cumulative count (i.e. t(g)) /
• for (g 0 g lt IM_DEPTH g)
• cumCountg histogramg prevCumCount
• prevCumCount cumCountg
• Exercise calculate new pixel grey-level using
  mapping function given earlier

21
Summary

• Histogram counts of pixels at each pixel
  amplitude
• useful for some transformations
• Contrast enhancement
• based on global or local image information
• pixel amplitude spread over wider range
• histogram stretch
• fundamentally a scaling operation
• combined with shift to counter side-effect
• histogram equalisation
• tries to obtain flatter histogram