Back to Adaptive Vision Library website

You are here: Start » Function Reference » Image Features » ImageLocalMaxima

ImageLocalMaxima


Finds image locations characterized by locally maximal pixel values.

Syntax

C++
C#
 
void avl::ImageLocalMaxima
(
	const avl::Image& inImage,
	atl::Optional<const avl::Region&> inRoi,
	bool inConsiderPlateaus,
	atl::Optional<float> inMinValue,
	atl::Optional<float> inMaxValue,
	float inMinDistance,
	atl::Optional<const avl::ImageLocalExtremaVerification&> inMaximaVerification,
	atl::Optional<atl::Array<avl::Region>&> outMaximaRegions,
	atl::Optional<atl::Array<avl::Point2D>&> outMaximaPoints,
	atl::Optional<atl::Array<float>&> outMaximaValues
)

Parameters

Name Type Range Default Description
inImage const Image& Input image
inRoi Optional<const Region&> NIL Range of pixels to be processed
inConsiderPlateaus bool Consider multi-pixel maxima (plateaus) or not
inMinValue Optional<float> NIL Minimal value of maximum to be considered
inMaxValue Optional<float> NIL Maximal value of minimum to be considered
inMinDistance float 0.0 - Minimal distance between two found maxima
inMaximaVerification Optional<const ImageLocalExtremaVerification&> NIL Maxima verification structure
outMaximaRegions Optional<Array<Region>&> Regions of local maxima (plateaus and singletons)
outMaximaPoints Optional<Array<Point2D>&> Maxima points and centers of plateaus, if requested
outMaximaValues Optional<Array<float>&> Values of maxima

Description

The operation finds local maxima on the image inImage within specific region inRoi.

It returns three arrays of the same length:

  • Subpixel precise positions of the maxima
  • Values of the maxima
  • Regions of equally bright pixels around the maxima

If the given image consists of more than one channel, they are averaged to obtain monochromatic image, so that the values of the extrema are of type float.

There are two types of local extrema:

  • Peak/hole: single pixel (strictly) brighter/darker than its neighbourhood (eight pixels). The extremum's position is determined with subpixel precision using gradient method with Sobel gradient. Returned region contains only one pixel.
  • Plateau/lowland: a connected set of equally bright pixels (strictly) brighter/darker from their neighbourhood. Returned position for this extremum is the center of mass of the plateau.

If parameter inConsiderPlateaus is set to true, the filter finds both types. Otherwise, it finds only strict maxima.

Parameter inMinValue determines the minimal value of the maximum to be considered.

Hints

  • If some points are not detected, try decreasing inMinValue or setting inConsiderPlateaus to True.
  • Consider adding some Gaussian smoothing before this function.

Examples

Example image

Output for example image

Remarks

If a plateau/lowland is concave, it's center of mass may lie outside it. In such case, the value of the extremum (which is equal to the common value of its pixels) may be different from the value of pixel indicated by position of the extremum.

See Also

ImageLocalMinima