Threshold Image Filter
Every pixel goes pure black or pure white: drag the slider to pick where the line falls.
or browse filesdrop an image here or paste
.png
Every pixel goes pure black or pure white: drag the slider to pick where the line falls.
or browse filesdrop an image here or paste
A threshold filter makes one decision about every pixel: is it brighter than the level you set, or darker? Brighter goes pure white, darker goes pure black, and nothing survives in between. There are no greys in the result at all, which is what separates thresholding from an ordinary black and white conversion. Drag the slider and the preview follows, so you can watch text detach from a grey page or a face collapse into a stencil and stop exactly where it looks right. Or press Auto and let the page pick the level for you.
People look for this as a threshold generator, a threshold maker or a threshold editor; they are the same one-slider operation, and it runs on photos, scans and screenshots alike. The level runs from 0 to 255 and starts at 128, the middle. A pixel whose brightness is at or above the level turns white, anything below it turns black. Lower it and only the darkest parts stay black; raise it and more of the picture falls to black. On a scan of a document the useful range is usually narrow, so move in small steps.
Drop a photo or a scan onto the page, or paste one from the clipboard, then drag the slider until the result reads the way you want. Download as PNG, which is the right format for flat black and white images and keeps the file small, or as JPG or WebP if something downstream needs them. Everything runs in your browser: the image is never uploaded, there is no signup, and no watermark is added.
Global mode uses one level for the whole image, the classic threshold effect. Auto, next to the slider, finds that level for you with Otsu's method: it looks at the histogram of the picture and splits it where the dark group and the light group separate most cleanly. On a scan or a logo it usually lands right first time, and you can still nudge the slider afterwards.
Adaptive mode is the answer to a page photographed under uneven light. Instead of one number, every pixel is compared with the average brightness of its own neighbourhood, so a shadow across one corner no longer fills that corner with black. The slider becomes Bias there: raise it to clean up a speckled background, lower it to make thin strokes bolder. This adaptive threshold is what document scanners and OCR software use before reading text.
The Invert switch swaps black and white, for white lines on a black background or a negative stencil.
Scanned documents and phone photos of paper are the most common reason to binarize an image. Uneven lighting leaves the page grey and the text muddy; a threshold throws the page away and keeps the ink, which is what OCR software wants to see and what makes a scan print cleanly.
Beyond documents, a binary image is the input that cutting and printing machines expect. Stencils, screen-printing separations, vinyl cutters, laser engravers and embroidery digitisers all need a shape that is either on or off, not a gradient. The same conversion is a strong graphic effect in its own right for posters, stickers and high-contrast portraits.
There is no correct number, only the one that suits your image, and Auto is a good place to start. For dark text on white paper, raise the level if letters break up into dots, and lower it as soon as the background starts to speckle. For a portrait, higher levels keep more shadow and read as a woodcut; lower levels leave only the darkest features and read as a stencil.
If the photo is lit unevenly, no single level will work everywhere at once - one side of the page will fill in before the other clears. Switch to Adaptive, which handles exactly that. For a photo rather than a page, fixing the lighting first helps: raise the contrast or the brightness before thresholding, and the level that works becomes much easier to find.
They are different conversions and it is worth knowing which you want. The black and white converter is for making a photo black and white: it keeps the full range of greys, so the photograph still looks like a photograph, and it has a quick No grey mode for when that is all you need. A threshold keeps two values only, and this page is for when the cut matters - an exact level, Auto, or Adaptive for a scan with a shadow across it. If your result looks harsh and you wanted a normal monochrome photo, the converter is the tool; if it looks too soft and you wanted flat shapes, you are in the right place.
The preview and the download apply the same cut to the pixels, so what you see is what you save, at the full resolution of the file you loaded. Transparency is kept: a PNG with a transparent background comes back thresholded with the background still transparent, rather than gaining a black one, and saved as JPG it sits on white. All of it happens on your device through the canvas API, so no image is transmitted anywhere and the page keeps working offline once it has loaded.
Open the picture here and drag the slider, or press Auto. Every pixel at or above the level becomes white and every darker pixel becomes black, and the preview updates as you drag. Download it as PNG, JPG or WebP at the original resolution. No account or upload is involved.
That is what this page does - binarization and thresholding are the same operation. Load the image, pick the level or press Auto, and the result contains only two values, 0 and 255, with no greys between them.
A black and white conversion keeps every shade of grey, so a photo still looks like a photo. A threshold allows only pure black and pure white, so the result is flat shapes. Use the threshold when something downstream needs an on-or-off image, such as OCR, a stencil or a cutting machine.
It depends on the image. Auto picks one with Otsu's method, which suits most scans and logos; from there, portraits often want a higher level to keep some shadow. Move in small steps and watch the point where the background starts to speckle - that is one step too far.
Because the page was lit unevenly, so one area is genuinely darker than another and a single level cannot suit both. Switch to Adaptive mode, which compares every pixel with its own surroundings instead of one number for the whole page, and the shadowed corner comes out as clean as the rest.
No. The picture is decoded, thresholded and encoded in your browser, at preview and at full download resolution alike. Nothing is sent to any server, and the tool keeps working with the network disconnected once the page is open.