dual_threshold

dual_threshold ( Image : RegionCrossings : MinSize, MinGray, Threshold : )

Threshold operator for signed images.

dual_threshold segments the input image into a region with gray values >= Threshold (``positive'' regions) and a region with gray values <= -Threshold (``negative'' regions). ``Positive'' or ``negative'' regions having a size of less than MinSize are suppressed, as well as regions whose maximum gray value is less than MinGray in absolute value.

The segmentation performed is not complete, i.e., the ``positive'' and ``negative'' regions together do not necessarily cover the entire image: Areas with a gray value between -Threshold and Threshold, -MinGray and MinGray respectively, are not taken into account.

dual_threshold is usually called after applying a Laplace operator (laplace, derivate_gauss or diff_of_gauss) or the difference of two images (sub_image) to an image.

The zero crossings of a Laplace image correspond to edges in an image, and are the separating regions of the ``positive'' and ``negative'' regions in the Laplace image. They can be determined by calling dual_threshold with Threshold = 1. The parameter MinGray controls the noise invariance, while MinSize controls the resolution of the edge detection.

Using byte images only the positive part of the operator is applied. Therefore dual_threshold behaves like a standard threshold operator (threshold) with successive connection and select_gray.

Parameters

`Image` (input_object)	image(-array) `->` object : byte / int2 / int4 / real
Input Laplace image.

`RegionCrossings` (output_object)	region-array `->` object
``Positive'' and ``negative'' regions.

`MinSize` (input_control)	integer `->` integer
Regions smaller than MinSize are suppressed.
Default value: 20
Suggested values: 0, 10, 20, 50, 100, 200, 500, 1000
Typical range of values: `0 <= MinSize <= 10000` (lin)
Minimum increment: 1
Recommended increment: 10

`MinGray` (input_control)	real `->` real
Regions whose maximum absolute gray value is smaller than MinGray are suppressed.
Default value: 5.0
Suggested values: 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 9.0, 11.0, 15.0, 20.0
Typical range of values: `0.001 <= MinGray <= 10000.0` (lin)
Minimum increment: 1.0
Recommended increment: 10.0
Restriction: `MinGray > 0`

`Threshold` (input_control)	real `->` real
Regions which have a gray value larger than Threshold (or smaller than -Threshold) are suppressed.
Default value: 2.0
Suggested values: 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 9.0, 11.0, 15.0, 20.0
Typical range of values: `0.001 <= Threshold <= 10000.0` (lin)
Minimum increment: 1.0
Recommended increment: 10.0
Restriction: `(Threshold >= 1) && (Threshold <= MinGray)`

Example

/* Edge detection with the Laplace operator (and edge thinning) */
diff_of_gauss(Image,Laplace,2.0,1.6) 
/* find "`positive"' and "`negative"' regions: */
dual_threshold(Laplace,Region,20,2,1) 
/*The zero runnings are the complement to these image section: */
complement('full',Region,Nulldurchgaenge).

/* Simulation of \OpRef{dual_threshold} */
dual_threshold(Laplace,Result,MinS,MinG,Threshold):
  threshold(Laplace,Tmp1,Threshold,999999) 
  connection(Tmp1,Tmp2) 
  select_shape(Tmp2,Tmp3,'area','and',MinS,999999) 
  select_gray(Laplace,Tmp3,Tmp4,'max','and',MinG,999999) 
  threshold(Laplace,Tmp5,-999999,-Threshold) 
  connection(Tmp5,Tmp6) 
  select_shape(Tmp6,Tmp7,'area','and',MinS,999999) 
  select_gray(Laplace,Tmp7,Tmp8,'min','and',-999999,-MinG) 
  concat_obj(Tmp4,Tmp8,Result).

Result

dual_threshold returns 2 (H_MSG_TRUE) if all parameters are correct. The behavior with respect to the input images and output regions can be determined by setting the values of the flags 'no_object_result', 'empty_region_result', and 'store_empty_region' with set_system. If necessary, an exception is raised.

Parallelization Information

dual_threshold is reentrant and automatically parallelized (on tuple level).

Possible Predecessors

min_max_gray, sobel_amp, gauss_image, reduce_domain, diff_of_gauss, sub_image, derivate_gauss

Possible Successors

connection, dilation1, erosion1, opening, closing, rank_region, shape_trans, skeleton

Alternatives

threshold, connection, select_shape, select_gray, dyn_threshold, check_difference

Module

Region processing