Choose a correlation resize factor rounded to an int

tests/unit_tests/test_scan_workflows.py

@@ -431,3 +431,33 @@ def test_histo_workflow_settings_ui(histo_workflow, mocker):
         "equal_distances",
     ]
     assert names == expected_names
+
+
+@pytest.mark.parametrize(
+    ("save_res,expected_resize",)[
+        ((1400, 1400), 1 / 2),  # width=1400 -> N=2
+        ((1640, 1232), 1 / 2),  # width=1640 -> round(1640/750)=2
+        ((760, 750), 1 / 1),  # width=760 -> N=1
+        ((1499, 1000), 1 / 2),  # width ~1500 -> N=2
+        ((2250, 1800), 1 / 3),  # width=2250 -> N=3
+        ((700, 700), 1 / 1),  # width < 750 -> N=1
+    ]
+)
+def test_correlation_resize(histo_workflow, save_res, expected_resize):
+    """Test that scan_workflows chooses a suitable correlation_resize factor.
+
+    correlation_resize is always a unit fraction (1/N, N integer) so that the
+    downsampled image has an area roughly equal to TARGET_STITCHING_DIMENSION**2.
+    This ensures stitching correlations are fast, robust, and avoid artefacts.
+
+    The value is found by taking the image area (width * height), dividing by the
+    target area, taking the square root, and rounding to the nearest integer N. Then
+    correlation_resize = 1 / N.
+    """
+    histo_workflow.save_resolution = save_res
+    histo_workflow.overlap = 0.1
+    settings = histo_workflow._get_stitching_settings_model()
+
+    assert isinstance(settings, StitchingSettings)
+    assert settings.correlation_resize == expected_resize
+    assert 0 < settings.correlation_resize <= 1