BoofCV / Java - Constructing Hessian matrix of an image with Gaussian convolution

Question

I want to build a Hessian matrix of a grayscale image but I need to calculate the second order derivatives. I'm working with BoofCV and I found out a function that calculates the first order:

ImageGradient_Gaussian<ImageUInt8,ImageFloat32> gaussian1 = new ImageGradient_Gaussian<ImageUInt8,ImageFloat32>(sigma, -1, ImageUInt8.class, ImageFloat32.class);
gaussian1.process(grayscaleImage, derivX, derivY);

I'm using as an input parameter "sigma" (standard deviation). Is there a function, or an other way, to calculate the second order derivatives using as input sigma and a grayscale image?

I'm open to other Java CV library suggestions.

score 1 · Accepted Answer · answered Jul 13 '15 at 03:05

1

You want to do something like:

GImageDerivativeOps.sobel(grey, derivX, derivY, BorderType.EXTENDED);
GImageDerivativeOps.hessianSobel(derivX, derivY, derivXX, derivXY, derivYY, BorderType.EXTENDED);

grey can be ImageFloat32 or ImageUInt8 deriv can be ImageFloat32 or ImageSInt16

The latest SNAPSHOT has an example, but it should work on older versions too:

https://github.com/lessthanoptimal/BoofCV/blob/master/examples/src/boofcv/examples/imageprocessing/ExampleImageDerivative.java

answered Jul 13 '15 at 03:05

lessthanoptimal

2,722
2
23
25

Thanks for the response. – Zerthimon Jul 13 '15 at 15:17
I was using a Gaussian kernel for first order partial derivatives... So hessianThree() method worked just fine for me! Thanks again! – Zerthimon Jul 13 '15 at 15:24

BoofCV / Java - Constructing Hessian matrix of an image with Gaussian convolution

1 Answers1