torchmorph

GPU-accelerated morphological image processing operations for PyTorch.

Installation

pip install torchmorph

Quick Start

import torch
import torchmorph as tm

x = torch.zeros(1, 1, 64, 64, device='cuda')
x[0, 0, 10:20, 10:20] = 1

dist = tm.euclidean_distance_transform(x)
dilated = tm.binary_dilation(x)

Spatial morphology and distance-transform functions accept CUDA tensors shaped as (B, C, Spatial...), with one to eight spatial dimensions. Binary operators treat nonzero values as foreground.

Optimal transport works on CPU and CUDA with batches of flattened histograms:

source = torch.rand(8, 64, device="cuda")
target = torch.rand(8, 64, device="cuda")

solver = tm.SinkhornSolver(epsilon=1.0, max_iter=200)
cost = tm.build_cost_matrix((8, 8), device=source.device)
distances = solver(source, target, cost)

API Reference

• Distance Transforms — Euclidean, Chamfer, and brute-force distance transforms
• Structuring Elements — Multidimensional connectivity structures
• Binary Morphology — Binary dilation, erosion, filling, and related operators
• Grayscale Morphology — Grayscale dilation, erosion, gradients, and top-hats
• Optimal Transport — Cost matrices and differentiable Sinkhorn transport