Example Thresholding

• Variable Lighting
• 

  Calibration grid. Left: original, Middle: Global Otsu, Right: Adaptive Square

• Difficult Text Example
• 

  Left: Original, Middle: Adaptive Square, Right: Sauvola

Thresholding gray scale images is one of the most basic ways to segment an image. It is quick and effective in many situations. BoofCV provides several algorithms for computing both global and local (adaptive) thresholds.

Example Code:

• ExampleThresholding.java

Concepts:

• Segmentation
• Thresholding

Relevant Examples/Tutorials:

• Tutorial Image Segmentation
• Example Binary Image
• Example Superpixels
• Example Color Segmentation

Example Code

/**
 * Demonstration of different techniques for automatic thresholding an image to create a binary image.  The binary
 * image can then be used for shape analysis and other applications.  Global methods apply the same threshold
 * to the entire image.  Local/adaptive methods compute a local threshold around each pixel and can handle uneven
 * lighting, but produce noisy results in regions with uniform lighting.
 *
 * @see boofcv.examples.imageprocessing.ExampleBinaryOps
 *
 * @author Peter Abeles
 */
public class ExampleThresholding {

	public static void threshold( String imageName ) {
		BufferedImage image = UtilImageIO.loadImage(imageName);

		// convert into a usable format
		ImageFloat32 input = ConvertBufferedImage.convertFromSingle(image, null, ImageFloat32.class);
		ImageUInt8 binary = new ImageUInt8(input.width,input.height);

		// Display multiple images in the same window
		ListDisplayPanel gui = new ListDisplayPanel();

		// Global Methods
		GThresholdImageOps.threshold(input, binary, ImageStatistics.mean(input), true);
		gui.addImage(VisualizeBinaryData.renderBinary(binary, null),"Global: Mean");
		GThresholdImageOps.threshold(input, binary, GThresholdImageOps.computeOtsu(input, 0, 256), true);
		gui.addImage(VisualizeBinaryData.renderBinary(binary, null),"Global: Otsu");
		GThresholdImageOps.threshold(input, binary, GThresholdImageOps.computeEntropy(input, 0, 256), true);
		gui.addImage(VisualizeBinaryData.renderBinary(binary, null),"Global: Entropy");

		// Local method
		GThresholdImageOps.adaptiveSquare(input, binary, 28, 0, true, null, null);
		gui.addImage(VisualizeBinaryData.renderBinary(binary, null),"Adaptive: Square");
		GThresholdImageOps.adaptiveGaussian(input, binary, 42, 0, true, null, null);
		gui.addImage(VisualizeBinaryData.renderBinary(binary, null),"Adaptive: Gaussian");
		GThresholdImageOps.adaptiveSauvola(input, binary, 5, 0.30f, true);
		gui.addImage(VisualizeBinaryData.renderBinary(binary, null),"Adaptive: Sauvola");

		// Sauvola is tuned for text image.  Change radius to make it run better in others.

		// Show the image image for reference
		gui.addImage(ConvertBufferedImage.convertTo(input,null),"Input Image");

		String fileName =  imageName.substring(imageName.lastIndexOf('/')+1);
		ShowImages.showWindow(gui,fileName);
	}

	public static void main(String[] args) {
		// example in which global thresholding works best
		threshold("../data/applet/particles01.jpg");
		// example in which adaptive/local thresholding works best
		threshold("../data/applet/segment/uneven_lighting_squares.jpg");
		// hand written text with non-uniform stained background
		threshold("../data/applet/segment/stained_handwriting.jpg");
	}
}