Example Image Convert

From BoofCV
Jump to navigationJump to search

Demonstrates how to convert between different types of BoofCV image types. When converting images into different types one needs to make sure the image which is being written into can store the full range of values of the original image, otherwise the output will be garbage. It is also demonstrated how to visualize a signed image using a built in function.

Example Code:

Concepts:

  • Images
  • Visualization

Example Code

/**
 * Demonstrates how to convert between different BoofCV image types.
 *
 * @author Peter Abeles
 */
public class ExampleImageConvert {
	// image loaded from a file
	BufferedImage image;
	// gray scale image with element values from 0 to 255
	GrayU8 gray;
	// Derivative of gray image. Elements are 16-bit signed integers
	GrayS16 derivX, derivY;

	void convert() {
		// Converting between BoofCV image types is easy with ConvertImage. ConvertImage copies
		// the value of a pixel in one image into another image. When doing so you need to take
		// in account the storage capabilities of these different class types.

		// Going from an unsigned 8-bit image to unsigned 16-bit image is no problem
		var imageU16 = new GrayU16(gray.width, gray.height);
		ConvertImage.convert(gray, imageU16);

		// You can convert back into the 8-bit image from the 16-bit image with no problem
		// in this situation because imageU16 does not use the full range of 16-bit values
		ConvertImage.convert(imageU16, gray);

		// Here is an example where you over flow the image after converting
		// There won't be an exception or any error messages but the output image will be corrupted
		var imageBad = new GrayU8(derivX.width, derivX.height);
		ConvertImage.convert(derivX, imageBad);

		// One way to get around this problem rescale and adjust the pixel values so that they
		// will be within a valid range.
		var scaledAbs = new GrayS16(derivX.width, derivX.height);
		GPixelMath.abs(derivX, scaledAbs);
		GPixelMath.multiply(scaledAbs, 255.0/ImageStatistics.max(scaledAbs), scaledAbs);

		// If you just want to see the values of a 16-bit image there are built in utility functions
		// for visualizing their values too
		BufferedImage colorX = VisualizeImageData.colorizeSign(derivX, null, -1);

		// Let's see what all the bad image looks like
		// ConvertBufferedImage is similar to ImageConvert in that it does a direct coversion with out
		// adjusting the pixel's value
		var outBad = new BufferedImage(imageBad.width, imageBad.height, BufferedImage.TYPE_INT_RGB);
		var outScaled = new BufferedImage(imageBad.width, imageBad.height, BufferedImage.TYPE_INT_RGB);

		var panel = new ListDisplayPanel();
		panel.addImage(ConvertBufferedImage.convertTo(scaledAbs, outScaled), "Scaled");
		panel.addImage(colorX, "Visualized");
		panel.addImage(ConvertBufferedImage.convertTo(imageBad, outBad), "Bad");
		ShowImages.showWindow(panel, "Image Convert", true);
	}

	/**
	 * Load and generate images
	 */
	public void createImages() {
		image = UtilImageIO.loadImageNotNull(UtilIO.pathExample("standard/barbara.jpg"));

		gray = ConvertBufferedImage.convertFromSingle(image, null, GrayU8.class);
		derivX = GeneralizedImageOps.createSingleBand(GrayS16.class, gray.getWidth(), gray.getHeight());
		derivY = GeneralizedImageOps.createSingleBand(GrayS16.class, gray.getWidth(), gray.getHeight());

		GImageDerivativeOps.gradient(DerivativeType.SOBEL, gray, derivX, derivY, BorderType.EXTENDED);
	}

	public static void main( String[] args ) {
		var app = new ExampleImageConvert();
		app.createImages();
		app.convert();
	}
}