Replace peak-magnitude scaling so absolute motion stays comparable across frames when saving RGB visualizations

[huidongc]

Description

Replace peak-magnitude scaling so absolute motion stays comparable across frames when saving RGB visualizations.

Fixes NVBUG#6520028

Type of change

• Bug fix (non-breaking change which fixes an issue)

Screenshots

Please attach before and after screenshots of the change if applicable.

Before	After

Checklist

• I have read and understood the contribution guidelines
• I have run the pre-commit checks with ./isaaclab.sh --format
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• I have added a changelog fragment under source/<pkg>/changelog.d/ for every touched package (do not edit CHANGELOG.rst or bump extension.toml — CI handles that)
• I have added my name to the CONTRIBUTORS.md or my name already exists there

[isaaclab-review-bot]

Isaac Lab Review Bot

The motion-vector display path now uses fixed [-1, 1] clamping instead of per-frame peak normalization, preserving comparable absolute motion across saved frames. The implementation is focused and compatible, but this user-visible change needs an Isaac Lab changelog fragment before merge.

• Design and architecture: The visualization policy remains localized in normalize_camera_output_for_display. Fixed scaling deliberately trades visibility of very small motion for consistent values across frames, matching the PR’s stated goal.
• API: The function signature, float32 [0, 1] output range, and three-channel layout remain unchanged, so existing grid and image-saving consumers stay compatible. Motion-vector pixel values change visibly, requiring release-note coverage under the repository rules.
• Implementation: Clamping, remapping, and appending a zero blue channel are shape-preserving and avoid the former division guard. Add a past-tense changelog fragment under source/isaaclab/changelog.d/ describing the fixed [-1, 1] motion-vector visualization scaling.

Minor fixes needed. Posted 1 actionable finding inline.

Automated review; human maintainers own approval decisions.

[greptile-apps]

Greptile Summary

The PR replaces frame-relative peak-magnitude scaling with fixed-range clamping so motion-vector colors remain comparable across frames.

• Clamps the U/V channels to [-1, 1], remaps them to RGB display values, and preserves a zero blue channel.
• Updates the unit-test expectation and changelog to reflect saturation outside the fixed range.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains; the previously reported mismatch between the clamp behavior and unit-test expectation is fixed at current HEAD.

Important Files Changed

Filename	Overview
source/isaaclab/isaaclab/utils/images.py	Replaces peak-based motion-vector normalization with fixed clamping and RGB packing.
source/isaaclab/test/utils/test_images.py	Correctly updates the motion-vector fixture expectation to match the new clamping behavior.
source/isaaclab/changelog.d/huidongc-fix-noise-in-motion-vectors.rst	Accurately documents the visualization behavior change and saturation semantics.

_{Reviews (2): Last reviewed commit: "Update test expectation" | Re-trigger Greptile}

[huidongc]

@greptile-apps

[huidongc]

@isaaclab-review-bot

[huidongc]

The franka cloth env was recently re-configured by #6831 to have near-zero gravity at beginning and linearly increased to 9.8m/s/s, therefore we can't see motion of the cloth any more in the first couple of frames.

Maybe we should make another PR to modify the franka cloth rendering test so that it is able to capture potential regression for deformable motion vectors.