Appendices - Feature Extraction and Image Processing - page 319

So let's work out the histogram of our eye image:

eye_histogram:=histogram(eye)

To display it, we need a horizontal axis which gives the range of brightness levels

bright:=0..255

400

eye _histogram bright

200

0 0

100 200

Bright

So here's the histogram of our picture of the eye image, p . The bright pixels relate mainly
to the skin , the darker ones to the hair .
The most common point operator replaces each pixel by a scaled version of the original
value. We therefore multiply each pixel by a number (like a gain ), by specifying a function
scale which is fed the picture and the gain, or a level shift (upwards or downwards). The
function scale takes a picture pic and multiplies it by gain and adds a level

scale(pic,gain,level):= for x ∈ 0..cols(pic)-1

Address the whole picture

for y ∈ 0..rows(pic)-1
newpic y,x ←

floor Multiply pixel

(gain·pic y,x +level) by gain and add level

Output the picture

newpic

So let's apply it:

brighter:=scale(eye, 1.2, 10)

You can change the settings of the parameters to see their effect, that's why you've got
this electronic document. Try making it brighter and darker. What happens when the gain
is too big (>1.23)?
So our new picture looks like the one overleaf (using Mathcad's picture display facility):
The difference is clear in the magnitude of the pixels, those in the 'brighter' image are
much larger than those in the original image, as well as by comparison of the processed
with the original image. The difference between the images is much clearer when we look
at the histogram of the brighter image. So let's have a look at our scaled picture:
b_eye_hist:=histogram(brighter)