LCE (local contrast enhancement) uses a process similar to histogram equalization (but with greater latitude for selecting the form of the output distribution) and adds in a term based on the deviation from a local estimate of the mean.
Mathematically, the output at a point, o(x,y), is equal to
min(m(x,y), mmax)
o(x,y) = c(x,y) * g(m(x,y)) + integral f(i) di
i = mmin
where m(x,y) is a local estimate of the mean intensity at (x, y) and c(x,y) is the difference between the intensity at (x,y) and the estimated local mean. The graph of the local mean stretch function shows the form of the function f; the horizontal axis is the local mean intensity and the vertical axis ranges from 0 to 1. The graph of the local contrast weighting function shows the form of the function g; the horizontal axis is again the local mean intensity and the vertical axis ranges from 0 to 2. In both graphs, a histogram of the local means for either current section (if the input is a window) or for the first section in the selected input region (if the input is a file) is shown as a reference. mmin and mmax are the minimum and maximum values on the horizontal axes; they can be specified in the MinMax field and when the autorange toggle is on they are automatically set based on the minimum and maximum local means for the section processed.
The graphs of the functions each have control points, shown as red squares, for modifying their shape. To use a control point, position the mouse pointer over it and press and hold the left mouse button. Drag the mouse until the control point is in the desired position and then release the left mouse button. The control points at the ends move vertically and horizontally; the control point in the middle of the mean stretch function only moves horizontally.
As a guide for choosing the form of the weighting functions here are a couple of the extreme cases:
Overview | Region processing | Box | Autorange | MinMax | Mean stretch function | Contrast weighting function
LCE 3D | Histogram equalization | Filter2D | Priism
The local mean intensity at a point (see the overview for how the local mean is used to calculate the enhanced result) is estimated by calculating the mean intensity for a square of n by n pixels about the point. n is the value shown in the Box field.
Overview | Region processing | Box | Autorange | MinMax | Mean stretch function | Contrast weighting function
When the AutoRange toggle is on, the range of values for which the weighting functions are calculated and plotted, is automatically set from the minimum and maximum local mean in the current section.
Overview | Region processing | Box | Autorange | MinMax | Mean stretch function | Contrast weighting function
The MinMax field shows the range of values on the horizontal axes of the local mean stretch graph and the local contrast weighting graph. When the AutoRange toggle is on, these bounds are automatically set based on the minimum and maximum local means for the current section (when processing is not in progress and the input is a file, the current section is the first section in the selected input region).
Overview | Region processing | Box | Autorange | MinMax | Mean stretch function | Contrast weighting function
As the overview describes in more detail, the result of local contrast enhancement at each pixel is the sum of two components: a contribution from the local mean of the input data about the point and a contribution from the local contrast (difference between the local mean and the point's value). The local mean stretch function affects the first and is plotted in red as a function of the local mean intensity. A histogram of the local mean values for the current section (if the input is an image window) or the first section in the input range (if the input is a file) is shown on the same axis for reference.
The curve has three control points to control the shape of the curve. To adjust a control point, position the mouse pointer over it, depress the right mouse button, drag the mouse, and then release the right mouse button. The control points on the left and right move both vertically and horizontally; the control point in the middle only moves horizontally.
Below the plot, the coordinates of the control points are shown and may be edited directly. The coordinates are normalized to the range of zero to one. The first two values shown are the horizontal and vertical coordinates, respectively, of the left control point. The third value is the horizontal coordinate of the middle control point. The fourth and fifth values are the horizontal and vertical coordinates, respectively, of the right control point.
Overview | Region processing | Box | Autorange | MinMax | Mean stretch function | Contrast weighting function
As the overview describes in more detail, the result of local contrast enhancement at each pixel is the sum of two components: a contribution from the local mean of the input data about the point and a contribution from the local contrast (difference between the local mean and the point's value). The local contrast weighting function affects the first and is plotted in red as a function of the local mean intensity. A histogram of the local means value for the current section (if the input is an image window) or the first section in the input range (if the input is a file) is shown on the same axis for reference.
The curve has two control points to control the shape of the curve. To adjust a control point, position the mouse pointer over it, depress the right mouse button, drag the mouse, and then release the right mouse button. The control points move both vertically and horizontally.
Below the plot, the coordinates of the control points are shown and may be edited directly. The coordinates are normalized to the range of zero to one. The first two values shown are the horizontal and vertical coordinates, respectively, of the left control point. The third and fourth values are the horizontal and vertical coordinates, respectively, of the right control point.
Overview | Region processing | Box | Autorange | MinMax | Mean stretch function | Contrast weighting function