{{ summaryTitle }}
{{ summaryValue }}

{{ summaryLine }}

{{ badge.label }} {{ badge.value }}
Drop, paste, or browse one PNG, JPEG, WebP, GIF, SVG, AVIF, BMP, or other browser-readable image up to 40 MB.
{{ sourceDropTitle }} {{ sourceDropLine }}

{{ sourceStatus }}

Choose the background evidence that best matches this image.
Pick a backdrop color, or load the current border estimate.
{{ sensitivity }}%
Raise for uneven paper, walls, and shadows; lower when subject colors resemble the background.
{{ edgeSoftness }} px
Use 0 px for hard artwork and about 2–4 px for many product-photo edges.
Choose a lower ceiling for a faster preview or a higher ceiling for more retained detail.
Switch mattes to reveal pale, dark, or semi-transparent edge residue.
{{ item.title }}{{ item.status }}
{{ item.detail }}
MeasureValueReview noteCopy
{{ row.label }}{{ row.value }}{{ row.note }}
Preparing coverage chart…

The chart renderer is unavailable. Exact coverage remains available in the mask ledger.

Image Metadata Stripper Strip hidden metadata from an image in your browser and create a clean JPEG, PNG, or WebP copy with a check for remaining privacy markers.
Image Compressor Compress one image locally in your browser, compare size and visual quality, then download a resized or reformatted copy with metadata controls.
Image Resizer Resize one image locally, compare fit or crop geometry, and tune format quality against a file-size target before downloading.
Image Format Converter Convert one image locally with format, resize, crop, quality, and transparency controls while checking the actual output and byte size.

Background removal turns selected pixels transparent so the remaining subject can be placed on another color or design. The result is a raster cutout, not a newly understood object. Its quality depends on whether color and border clues separate the subject from the backdrop.

A clean product photograph on paper, a logo on a flat color, or an object against a bright studio wall can work well with pixel rules. Hair, fur, glass, shadows, reflections, and busy rooms are harder because the boundary contains colors from both subject and background. Pale products on white and green details on a green screen are especially easy to over-remove.

Image situationUseful evidenceMain failure mode
Object on a plain sweepSimilar colors connected to the image borderShadows or subject edges join the backdrop
Logo on one flat colorOne chosen color wherever it appearsThe same color inside the logo disappears
Bright studio imageBright, low-chroma pixels connected to the borderWhite fabric, glass, or highlights are removed
Hair, fur, reflections, or a busy sceneNo single dependable color cluePatchy masks and damaged fine detail

Edge softness replaces a hard cut with partial transparency near the mask boundary. A little feathering can hide stair-stepped pixels, but too much creates a halo and can make a crisp logo look blurred. Transparency should therefore be inspected on both light and dark backgrounds before the cutout is used.

The output dimensions also matter. Large images may be reduced to a pixel ceiling before masking, which lowers memory use and speeds up processing but discards source resolution. A good-looking preview does not guarantee that tiny details survived the reduction.

Color-based masking is best treated as a fast method for simple backdrops. A low-confidence or visibly damaged edge calls for a manual mask or a subject-aware editor rather than more aggressive sensitivity.

How to Use This Tool:

Choose the mask evidence that matches the image, then judge the transparent edge on more than one matte.

  1. Upload, drop, or paste one browser-readable image up to 40 MB.
  2. Choose Edge-sampled backdrop for a plain background connected to the border, Picked color key for one flat color anywhere in the image, or Light studio backdrop for a bright, low-chroma border.
  3. Set the Background color when using a color key, then tune Sensitivity and Edge softness. Raise sensitivity for uneven backdrop shades, lower it when subject colors resemble the background, and use 0 px softness for hard artwork or about 2 to 4 px for many product-photo edges.
  4. Choose the Render ceiling. The 1.5 MP setting favors speed, 3 MP is the balanced default, and 6 MP retains more detail at higher memory cost.
  5. Inspect Transparent cutout and switch the Preview background between light, dark, and colored mattes. Download the PNG only after the edge and review flags agree with what you see.

Interpreting Results:

The fit score is a review aid, not an image-quality measurement. High begins at 76%, Review begins at 52%, and lower scores indicate Manual review. A high score can still accompany missing hair, transparent glass, or a clipped subject color.

  • Removed, Softened, and Retained should add to 100% of the rendered canvas.
  • Less than 1% removal or more than 92% removal raises a balance warning.
  • Border diversity above 38 raises a mixed-border warning for edge and light methods.
  • More than 16% softened pixels suggests that feathering affects too much of the image.
  • Every color-key result carries a review flag because matching color can also occur inside the subject.

Source pixels and generated output stay in browser memory, are excluded from the share URL, and are not uploaded or stored. The transparent PNG does not preserve animation, vector editability, or source metadata, and a large source can be downscaled before export.

Technical Details:

The transformation works on decoded RGBA pixels. RGB supplies the color evidence; alpha stores opacity. Pixels classified as background receive alpha 0, retained pixels keep their opacity, and eligible boundary pixels may receive an intermediate alpha value.

Transformation Core

MethodClassification ruleScope
Edge sampleRGB distance from the mean sampled border colorOnly matching pixels connected inward from an image edge
Color keyRGB distance from the chosen colorEvery matching pixel, connected or not
Light studioBright, low-chroma pixels or pixels close to the sampled border colorOnly matching pixels connected inward from an image edge

Edge and light masks start at the outer border and expand through four-way neighbors. A matching island enclosed by retained pixels is therefore left alone. Color keying has no connectivity test and can remove an enclosed subject area that shares the selected color.

Formula Core

Color distance is Euclidean distance in the decoded RGB coordinates. Sensitivity sets the acceptance threshold by a fixed multiplier.

d=(R-Rb)2+(G-Gb)2+(B-Bb)2 T=2.28S

S is the sensitivity value from 8 to 86, so the RGB-distance threshold ranges from 18.24 to 196.08. For the light method, luma is calculated with the BT.709 coefficients shown below, while chroma is the largest RGB channel minus the smallest.

Y′=0.2126R+0.7152G+0.0722B

Coverage percentages divide each pixel count by the rendered pixel count. The render scale is the smaller of rendered width divided by source width and rendered height divided by source height.

Rule Core

The fit score starts at 82 for edge or light masking and 72 for color keying. It decreases with border diversity, extreme removal, and broad softening; light masking also loses points when sampled border luma is below 205. A clean edge sample with diversity below 18 and removal from 4% to 85% gains 8 points. The final score is rounded and limited to 5% through 96%.

These rules describe confidence in the available pixel evidence. They do not detect people, products, logos, or semantic foreground objects.

Worked Examples:

Balanced edge-sampled cutout

A 640 by 400 image contains 256,000 rendered pixels. If 150,000 are removed, 6,000 are softened, and 100,000 are retained, coverage is about 58.6%, 2.3%, and 39.1%. With sensitivity 38, the RGB threshold is 86.64. A low-diversity border and moderate removal can produce a high fit score, but the cutout still needs inspection on contrasting mattes.

References: