Grid ROI: Missing spacing parameters for non-uniform grids

[PierreRaybaut]

Grid ROI: Missing spacing parameters for non-uniform grids

Description

The grid ROI feature is not working correctly for images where subimages/features don't fill the entire image area. This is particularly problematic for real-world data like:

• Laser spot arrays with gaps between spots
• Diffraction patterns
• Any regularly spaced features with significant gaps or offsets

Problem

Users can adjust:

• Number of rows/columns (nx, ny)
• ROI size (xsize, ysize) - percentage of cell size
• Position offset (xtranslation, ytranslation) - percentage translation

But cannot adjust:

• Spacing between ROIs (distance between adjacent ROI centers)

The grid ROI extraction always assumes an evenly distributed grid across the entire image, calculating cell spacing as width / nx and height / ny. This fails when:

1. Spots/features don't fill the whole image
2. There's significant offset from the edges
3. The actual spacing between features differs from the calculated cell width/height

Example Case

When trying to extract a 7×7 spot grid from laser_spot_array_raw.png:

• The spots have gaps between them
• The first spot has a significant offset from the image edges
• The automatically computed spacing (image_width / 7) doesn't match the actual spot spacing
• Result: ROIs don't align with the actual spots

Classification

This is a bug fix rather than a feature addition because:

• The grid ROI feature claims to extract grids of subimages
• It fails for a significant class of real-world images (non-uniform/gapped grids)
• The missing parameters are essential for the feature to work as intended
• Users have no workaround without modifying image coordinates or manually creating ROIs

Proposed Solution

Add two new parameters to ROIGridParam:

• xstep: Horizontal spacing between ROI centers, as percentage of auto-computed cell width (default 100%)
• ystep: Vertical spacing between ROI centers, as percentage of auto-computed cell height (default 100%)

With default values (100%), the behavior remains unchanged (evenly distributed grid). Users can adjust these values to:

• < 100%: Tighter spacing (ROIs closer together)
• > 100%: Wider spacing (ROIs farther apart)

Impact

Without this fix, the grid ROI feature is essentially unusable for a large class of real-world scientific data where features are regularly spaced but don't fill the entire image.

Related Code

• Sigima: sigima/proc/image/extraction.py - ROIGridParam class and generate_image_grid_roi() function
• DataLab: datalab/gui/roigrideditor.py - Grid ROI editor dialog

Affected Versions

• DataLab 1.0.2
• Sigima 1.0.3