ContourRenderer

Smooth contour rendering using marching squares.

Provides sub-pixel accurate contour extraction and visualization using skimage.measure.find_contours.

Classes

class ContourRenderer

Render segmentation contours with multiple visualization modes.

Supports two outline modes: - Smooth: Marching squares (skimage.measure.find_contours) for sub-pixel

accurate contour extraction. Shows the interpolated boundary.

• Pixel: Morphological dilation-erosion for jagged outlines that show the actual pixel boundaries of the segmentation.

Both modes have their uses: - Smooth: Better for seeing true boundary shape, comparing curves - Pixel: Shows actual voxel boundaries, important for understanding

what the algorithm actually segmented at pixel level

Methods:

__init__()

Initialize renderer.

get_morphological_outline()

Get pixel-level outline using morphological operations.

find_contours()

Find smooth sub-pixel contours using marching squares.

draw_contours()

Draw smooth contour polylines on RGB image.

create_comparison_image()

Create comparison image with both segmentations as outlines.

create_error_image()

Create error analysis image with TP/FP/FN coloring.

create_combined_image()

Create combined visualization with both contours and error regions.

clear_cache()

Clear the contour cache.

Functions

get_morphological_outline()

Get pixel-level outline using morphological operations.

Creates a jagged outline that shows the actual pixel boundaries of the segmentation (dilation - erosion).

Args:

mask: 2D binary mask. iterations: Thickness of outline in pixels.

Returns:

Binary mask of outline pixels.

find_contours()

Find smooth sub-pixel contours using marching squares.

Args:

mask: 2D binary mask (any dtype, will be converted to bool).

Returns:

List of contour arrays, each with shape (N, 2) in (row, col) format.

draw_contours()

Draw smooth contour polylines on RGB image.

Args:

rgb: RGB image (H, W, 3) as float64 or uint8. contours: List of contour arrays from find_contours. color: RGB color tuple (0-255). thickness: Line thickness in pixels.

Returns:

RGB image with contours drawn.

create_comparison_image()

Create comparison image with both segmentations as outlines.

Args:

image_slice: 2D grayscale image (uint8) trial_slice: 2D binary trial segmentation mask gold_slice: 2D binary gold standard mask trial_color: RGB color for trial contour (default: green) gold_color: RGB color for gold contour (default: magenta) line_thickness: Contour line thickness in pixels mode: Outline mode - “smooth” (marching squares) or “pixel” (morphological)

Returns:

RGB image as uint8 array (H, W, 3)

create_error_image()

Create error analysis image with TP/FP/FN coloring.

Color coding (assuming gold standard is truth): - Green: True Positive (both agree) - Red: False Positive (trial only - over-segmentation) - Blue: False Negative (gold only - under-segmentation)

Args:

image_slice: 2D grayscale image (uint8) trial_slice: 2D binary trial segmentation mask gold_slice: 2D binary gold standard mask tp_color: RGB for true positives (agreement) fp_color: RGB for false positives (over-segmentation) fn_color: RGB for false negatives (under-segmentation) alpha: Opacity of overlay

Returns:

RGB image as uint8 array (H, W, 3)

create_combined_image()

Create combined visualization with both contours and error regions.

Args:

image_slice: 2D grayscale image (uint8) trial_slice: 2D binary trial segmentation mask gold_slice: 2D binary gold standard mask show_contours: Whether to draw smooth contours show_error_regions: Whether to show TP/FP/FN regions error_alpha: Opacity of error regions

Returns:

RGB image as uint8 array (H, W, 3)

clear_cache()

Clear the contour cache.