Color Region Cutout & Mask

What it is for

A logo on a plain white background, a product shot against a colored backdrop, a scanned signature: when the part you want to remove is one even color, a magic-wand selection is often all you need. This tool does that for one uploaded image. You point at a color with percentage coordinates, choose how similar other colors must be, and get back a transparent PNG and a grayscale mask you can reuse in another editor.

The image is uploaded to the server. PNG, JPEG, WebP and the first frame of a GIF are accepted, up to 8 MB and 16 megapixels, and the working copy is reduced to at most 1024 pixels on its longest side. The tool matches colors only; it does not recognize people or objects. Lumina's Selection Lab runs the same selection engine locally and adds rectangles, ellipses and combined selections.

Worked example

We uploaded a 1600 × 600 PNG with the black word "Hamburg" on a white background and used Seed X 2 %, Seed Y 2 % (a white pixel near the top-left corner), tolerance 10, contiguous pixels only on, feather 2 and the result Keep selected area.

The tool reported 360075 pixels selected (91.57 %) on the 1024 × 384 working image, with bounds covering the whole image. The selection is the connected white background. Everything else, about 33 000 pixels, is the black lettering, its anti-aliased edges and the white counters inside letters such as "a", "b" and "g", which the background does not touch.

With Keep selected area, the PNG keeps the white background and the letters become transparent. Choose Erase selected area instead and the background disappears while the lettering stays; the selection and its counts are identical either way.

How to read the result

• Selected pixels and selected area count the hard selection before feathering.
• Bounds give the smallest rectangle around the selection in working-image pixels.
• Download color region PNG is the edited image with transparency.
• Download grayscale selection mask is opaque: white is selected, black is not, and gray marks feathered coverage.
• Both files use the working size, so a large photo comes back smaller than the original.

Tips and common mistakes

• Pick a seed inside a large flat area, not on an anti-aliased edge; the tolerance is measured from the seed color, not from each neighbor.
• Start with a low tolerance (10–30) and raise it if parts of the area stay unselected. Too high and the selection leaks into similar colors.
• Turn contiguous pixels only off to catch enclosed areas, such as the inside of an "o", or every matching color in the image.
• Invert selection turns "select the background" into "select the subject" without moving the seed.
• Feather softens both sides of the edge. Values of 1–3 hide jagged edges; large values create a glow.
• JPEG noise makes flat colors uneven. Use a PNG source when you can, or raise the tolerance slightly.

FAQ

Can it remove the background behind a person?

Only if that background is one fairly even color. It matches colors and does not detect subjects.

Why is my output smaller than the uploaded image?

The tool works on a copy that is at most 1024 pixels on its longest side and returns that size. Selection Lab uses the same limit but keeps the image on your device.

What does 100 % for Seed X mean?

It addresses the last column of pixels. 0 % is the first column; the same applies to Seed Y and rows.

Does the mask include the feathering?

Yes. The grayscale mask stores the feathered coverage, while the reported pixel counts describe the hard selection.