[Hendrik-code]

What

RandomRedistributeSegGPU takes min_redist_std (default 0.01, the hardcoded value it replaces): the floor on each label's Gaussian bump width, in the [0, 1] min-max units the remap is built in.

Why

The remap adds a signed Gaussian bump per label at that label's mean intensity, then rescales the sum to a peak-to-peak of 2. A bump of width s, rescaled that way, has a slope of about 1.2 / s. At s = 0.01 a thin band of intensities swings by more than the whole input range, so voxel noise inside that band comes out as salt-and-pepper.

Measured on an intensity ramp with the floor binding (voxel-random labels, std_noise_range=(0, 0)), mean of the maximum slope over 20 seeds:

floor	0.01	0.05	0.1
max slope (mean)	123	26	13
1 + 1.2 / floor	121	25	13

On real SpineGAN MR/CT patches, the high-frequency share of the image (std of what a σ=1 blur removes ÷ std) at least doubled in 24% of Redistribute applications, and in 47% when it ran after SynthSeg. It is the most frequent TA in the paper config (p: 0.4).

Behaviour

Default unchanged (tested: bit-identical with and without the key at 0.01). A non-positive floor is refused. all_augmentations.json regenerated.

Tests

unit_tests/test_redistribute_std_floor.py:

• the measured slope follows 1 + 1.2 / floor within ×0.5 to ×1.5 at three floors;
• the default is the historical floor;
• a non-positive floor is refused;
• the registry accepts the key.

Full suite: 477 passed, 13 skipped. mypy smauglab/ clean, pre-commit passes.

🤖 Generated with Claude Code