Appendices - Feature Extraction and Image Processing - page 324

400

n_hist bright

200

0

0 50 100 150 200 250

Bright

The histogram now occupies the whole available range, as required.
Histogram equalisation is a nonlinear histogram-stretching process. The equalised
histogram is a resampled cumulative histogram . We first work out the histogram of the
picture, then we work out the cumulative histogram. Finally, we resample the cumulative
histogram, giving a look-up table to map original intensity levels to the equalised ones.
The main difference between equalisation and normalisation is that in normalisation all
grey levels have the same 'weight': the process stretches the histogram to occupy the
available range. In equalisation , the histogram is resampled or manipulated, again to cover
the available range. Since the histogram is manipulated, brightness values do not have the
same weight.

equalise(pic):= range ←

Define output range

255

number ←

Number of points

rows(pic)·cols(pic)

for bright ∈ 0..255
pixels_at_level bright ←

Initialise histogram

0

for x ∈ 0..cols(pic)-1
for y ∈ 0..rows(pic)-1

pixels_at_level

pic

y,x

Determine histogram

pixes_at_level +1

pic

y,x

sum ← 0
for level ∈ 0..255
sum ← sum+pixels_at_level level Form cumulative histogram

range
number

hist floor sum

level

Make look-up table

+0.00001

for x 0..cols(pic)-1
for y 0..rows(pic)-1
newpic

Map input to output

hist

y,x pic y,x

newpic