I believe that they all do the same task of distributing the frequency of pixel intensities to a bigger range, but can these terms be used interchangeably?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
I believe that they all do the same task of distributing the frequency of pixel intensities to a bigger range
They do.
but can these terms be used interchangeably?
no, they can't. because the way they increase the dynamic range of the image diffears.
All this methods do a transformation of the pixel intensity $u$ of and image into a new pixel intensity $v$ $$v = f(u) $$ in order to improve the dynamic range of the image.
for constrast stretching (that as far as I know is a more general case of image normalization) the transformation is a piecewise linear function
$$ f(u) = \begin{cases} \alpha u & 0 \leq u < a \\ \beta(u - a) + v_a& a \leq u < b \\ \gamma(u - b) + v_b & b \leq u < L \end{cases}$$
Where L is the total number of gray levels.
On the other hand, Histogram equalization is a more powerful technique for image enhancement that aims to give the histogram of a given image the desired uniform shape. This is achieve by the transformation $$v_{aux} = F_u(u) = \sum_{i=0}^u p_u(i) $$
$$v = int\left( \frac{v_{aux} - v_{aux_{min}}}{1 - v_{aux_{min}}} (L-1) + 0.5 \right) $$
where $p_u(i) $ is the normalized value of the histogram at the grey level $i$
