Fill gaps between regions (depending on gray value or color) or split overlapping regions.
expand_gray closes gaps between the input regions, which resulted from the suppression of small regions in a segmentation operator, (mode 'image'), for example, or separates overlapping regions 'region'). Both uses result from the expansion of regions. The operator works by adding a one pixel wide ``strip'' to a region, in which the gray values or color are different from the gray values or color of neighboring pixles on the region's border by at most Threshold (in each channel). For images of type 'cyclic' (e.g., direction images), also points with a gray value difference of at least 255 - Threshold are added to the output region.
The expansion takes place only in regions, which are designated as not ``forbidden'' (parameter ForbiddenArea). The number of iterations is determined by the parameter Iterations. By passing 'maximal', expand_gray iterates until convergence, i.e., until no more changes occur. By passing 0 for this parameter, all non-overlapping regions are returned. The two modes of operation ('image' and 'region') are different in the following ways:
'image'
The input regions are expanded iteratively until they
touch another region or the image border, or the expansion
stops because of too high gray value differences. Because
expand_gray processes all regions simultaneously, gaps
between regions are distributed evenly to all regions with
a similar gray value. Overlapping regions are split by
distributing the area of overlap evenly to both regions.
'region'
No expansion of the input regions is performed. Instead,
only overlapping regions are split by distributing the
area of overlap evenly to regions having a matching gray
value or color.
Because regions are only expanded into areas having a matching gray value or color, usually gaps will remain between the output regions, i.e., the segmentation is not complete.
|
Regions (input_object) |
region(-array) -> object |
| Regions for which the gaps are to be closed, or which are to be separated. | |
|
Image (input_object) |
image -> object : byte / cyclic |
| Image (possibly multi-channel) for gray value or color comparison. | |
|
ForbiddenArea (input_object) |
region -> object |
| Regions in which no expansion takes place. | |
|
RegionExpand (output_object) |
region(-array) -> object |
| Expanded or separated regions. | |
|
Iterations (input_control) |
string -> string / integer |
| Number of iterations. | |
| Default value: 'maximal' | |
| Suggested values: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 'maximal' | |
| Typical range of values: 1 <= Iterations <= 500 (lin) | |
| Minimum increment: 1 | |
|
Recommended increment: 1 | |
|
Mode (input_control) |
string -> string |
| Expansion mode. | |
| Default value: 'image' | |
| List of values: 'image', 'region' | |
|
Threshold (input_control) |
integer(-array) -> integer |
| Maximum difference between the gray value or color at the region's border and a candidate for expansion. | |
| Default value: 32 | |
| Suggested values: 5, 10, 15, 20, 25, 30, 40, 50 | |
| Typical range of values: 1 <= Threshold <= 255 (lin) | |
| Minimum increment: 1 | |
|
Recommended increment: 5 | |
#include <iostream.h>
#include "HalconCpp.h"
int main (int argc, char *argv[])
{
HImage image (argv[1]);
HRegion empty_region;
HWindow win;
image.Display (win);
HRegionArray seg = (image >= 100).Connection ();
seg.Display (win);
HRegionArray exp = seg.ExpandGray1 (image, empty_region,
"maximal", "image", 32);
win.SetDraw ("margin");
win.SetColored (12);
exp.Display (win);
win.Click ();
return (0);
}
expand_gray always returns the value 2 (H_MSG_TRUE). The behavior in case of empty input (no regions given) can be set via set_system('no_object_result',<Result>), the behavior in case of an empty input region via set_system('empty_region_result',<Result>), and the behavior in case of an empty result region via set_system('store_empty_region',<true/false>). If necessary, an exception handling is raised.
expand_gray is reentrant and processed without parallelization.
connection, regiongrowing, pouring, class_ndim_norm
expand_gray_ref, expand_region
Region processing