Compare the marks you can see.

Both images use the same segmentation mode, cutoff, and minimum component size. Dark and light modes select pixels at or below / at or above the luminance cutoff. Local contrast selects pixels whose luminance differs from the 31 × 31 neighborhood mean by at least the cutoff; its neighborhood uses surrounding pixels even outside the ROI. The minimum is measured in pixels after each image is independently resized to a maximum dimension of 1,000 pixels, so it is not a physically calibrated mark size. Blue areas are retained 8-connected components; touching marks can merge and textured marks can split. The ROI border is red. Drag either image to redefine its region.

Coverage is retained dark-pixel area divided by ROI area. Candidate density is per 10,000 ROI pixels; median size is in analysis pixels. The orientation summary is area-weighted and undirected: 0° is horizontal, 90° vertical. Low concentration means the direction is diffuse or the components are near-round. Images may have different exposures, grounds and scales; shared settings alone do not make them directly comparable. The sensitivity test varies the cutoff across nine values (64–192 for luminance modes, 4–36 for local contrast) while keeping both regions and the minimum candidate size fixed. It shows whether the sign and magnitude of the coverage difference change with the cutoff; it is a robustness check, not a probability or a significance test. This is a spatial comparison, not a reconstruction of time or an inferential test.

Images are processed in your browser. Export includes the source filenames, original and analysis dimensions, regions, settings, summary statistics, and candidate geometry, but not the images.