| |
|
Fixed Filters
Fixed filters are actually ordinary linear convolution filters with fixed kernel size.
The size of kernel is denoted at the and of filter's name. Letter V at the end of name means vertical
while H means horizontal. Here is the list of fixed filters with their respective kernels.
| Prewitt 3x3V | Prewitt 3x3H | Sobel 3x3V |
|---|
| -1 0 1 | 1 1 1 | -1 0 1 |
| -1 0 1 | 0 0 0 | -2 0 2 |
| -1 0 1 | -1-1-1 | -1 0 1 |
| Sobel 3x3H | Laplace 3x3 | Laplace 5x5 |
|---|
| 1 2 1 | -1-1-1 | -1-3-4-3-1 |
| 0 0 0 | -1 8-1 | -3 0 6 0-3 |
| -1-2-1 | -1-1-1 | -4 6 20 6-4 |
| | | -3 0 6 0-3 |
| | | -1-3-4-3-1 |
| Gauss 3x3 | Gauss 5x5 | Hpass 3x3 |
|---|
| 1 2 1 | 2 7 12 7 2 | -1-1-1 |
| 2 4 2 | 7 31 52 31 7 | -1 8-1 |
| 1 2 1 | 15 52 127 52 15 | -1-1-1 |
| | 7 31 52 31 7 | |
| | 2 7 12 7 2 | |
| Hpass 5x5 | Sharpen 3x3 | |
|---|
| -1-1-1-1-1 | -1-1-1 | |
| -1-1-1-1-1 | -1 16-1 | |
| -1-1 24-1-1 | -1-1-1 | |
| -1-1-1-1-1 | | |
| -1-1-1-1-1 | | |
Nonlinear Filters
In this section we'll shortly describe non-linear filters. Those filters can have various kernel sizes,
as well as various anchor positions.
Size of kernel is regulated by X and Y edit fields or spinners.
Position of anchor point is regulated by Xc and Yc edit fields and spinners.
Median
filter founds pixel value that has the highist appearing frequency inside given kernel. When found, anchor pixel
is replaced by this value.
Maximum
filter searches for highist pixel value inside the given kernel. Anchor pixel is replaced by this value.
Minimum
filter searches for lowest pixel value inside the given kernel. Anchor pixel is replaced by this value.
Example of IPL filter usage with contour.
Original picture is on the left, right picture is processed with IPL filters.
Click on picture to enlarge.
|
|
Morphological Filters
Morphological operations are simple erosion and dilation of an image.
Number of erosions and dilations are controlled by Iterations edit field/spinner.
Generally speaking morphological operations are reflected on boundaries of objects inside the image.
Erode
operation removes spurious pixels (such as noise) and thins boundaries of objects on a dark background.
Dilate
operation fills up holes and thick boundaries of objects on a dark background.
Open
operation eliminates small and thin objects, breaks objects at thin points and performs smoothing the
boundaries of large objects, without significantly changing their area.
Close
operation fills small and thin holes in objects, connects nearby objects and performs smoothing the
boundaries of large objects, without significantly changing their area.
Applying IPL Filter
If you want to apply IPL Filter on whole image, after choosing the filter and setting up parameters,
simply press Image button.
Filter will be applied to whole image.
You can apply IPL Filter through brush movement, contour and mask as well. For shape selection and brush movement
refer to Shapes and Brushes section.
NOTE:
Resultant image is affected by all previously set
Global parameters.
All kernel elements that are beyond the image border are wrapped around if
Wrap is checked.
NOTE:
IPL Filters are extremely fast when program is installed on Intel Pentium processor family, especially
on MMX processors.
Depending on screen resolution, not all methods are visible in list box. List box can be scrolled up or down.
Example of IPL filters usage.
In the upper left corner is original picture, other pictures show usage of
IPL filters.
Click on picture to enlarge.
|