kornia.filters

Differentiable image filtering on batched (B, C, H, W) tensors. Every function has an nn.Module counterpart on the same page.

Blurring	Gaussian, box, median, bilateral, guided and motion blur; blur pooling; unsharp masking.
Edge detection	Sobel, Laplacian and Canny edge detectors and spatial gradients in 2D and 3D.
Thresholding	Range thresholding and Otsu’s threshold.
Filtering API	Apply your own 2D, separable or 3D kernels with filter2d(), filter2d_separable() and filter3d(), or by correlation or convolution with correlate2d(), convolve2d(), correlate3d() and convolve3d(); fft_conv() filters in the frequency domain.
Kernels	Gaussian, box, Hanning, Laplacian, gradient and motion kernels used by the filters above.