commonly used in computer vision applications to create a robust, efficient, generic basis set for applications such as texture synthesis (Olshausen, Simoncelli). These local features consisted of:
- Gaussian: the image blurred by Gaussian kernels of widths 1, 2, and 3 mm;
- Magnitude of the gradient of Gaussians: the magnitude of the first spatial derivative of the image blurred by Gaussian kernels of widths 1, 2, and 3 mm;
- Laplacian of Gaussian: the trace of the spatial second derivative of the image blurred by Gaussian kernels of widths 1, 2, and 3 mm;
- Frobenius Norm of the Hessian: the matrix norm of the spatial second derivative of the image blurred by Gaussian kernels of widths 1, 2, and 3 mm