Ellipse and circle contour detection produces incorrect results

[StefanoPierini-Codra]

Description

Ellipse contour detection returns completely wrong results, while polygon contour detection works correctly on the same image.

Steps to reproduce

1. Create a new image using Create > Blobs (default 512×512)
2. Apply preprocessing (e.g. binarization via the test plugin)
3. Run Analysis > Contour detection with:
   • Threshold: 0.5 (default)
   • Shape: Polygon → result is correct ✅
4. Run Analysis > Contour detection again with:
   • Threshold: 0.5 (default)
   • Shape: Ellipse → result is completely wrong ❌

Expected behavior

Detected ellipses should overlay the blobs in the image, matching the contours found by the polygon mode.

Actual behavior

Ellipse contours appear at incorrect positions, not matching the actual objects in the image.

Environment

• Sigima 1.1.2
• scikit-image 0.26.0
• Python 3.14

[StefanoPierini-Codra]

Detected bug

In sigima/tools/image/preprocessing.py, the functions fit_ellipse_model and fit_circle_model use scikit-image's EllipseModel and CircleModel.

scikit-image 0.26 introduced a new API with named attributes:

• model.center returns (row, col) = (y, x)
• model.axis_lengths returns (semi_row, semi_col) = (b, a)

The code was reading center[0] as xc and center[1] as yc, but it should be the reverse.

Still to fix

After fixing this behaviour, placeholders appears at the right place but with the wrong size, as shown in the following image:

[StefanoPierini-Codra]

Two other problems were detected:

• wrong formula for ellipsis axis (a minus sign was missing)
• a problem with the radius conversion that was not converted from pixels to physical units

Both of them have been fixed in Sigima related issue DataLab-Platform/Sigima#21.

[ThomasMalletCodra]

Merge in sigima release (1.1.3) : DataLab-Platform/Sigima#22