Assessments¶
Each section is the report of one experiment with the code that produced it.
- Aggregation window — the plugin run on four images with a growing structure-tensor window σ and a growing vector-field grid: orientation distributions, color surveys and vector fields at every size.
- Benchmarking — the orientation distribution of OrientationJ against five other tools, on the same images and the same masks.
- Compare gradients — the five gradients of the plugin measured against analytic ground truth, error against structure size.
- Operator — the same measurement in sixty lines of separable convolutions, with no transform at all, and what it costs in accuracy.
- Structure vs nematic — the structure tensor against the nematic tensor and the naive mean of angles, on a saturated-ring image with an analytic reference, under saturation, ramp, noise and window size.