Watermarks, Explained

After reading this you will know how a watermark's opacity, size, rotation and tiling trade off between protection and visibility, and how to pick settings that survive casual copying.

A watermark is a mark laid over an image to signal ownership or to make casual reuse annoying. It can be text (your name, a URL, "DRAFT") or a small logo. The tool paints that mark onto every pixel it covers, either once in a chosen spot or repeated across the whole frame as a tiled pattern.

Here is the hook. Suppose you post a 2000 by 1333 photo. A single small logo in the corner is trivial to crop out: one rectangular selection and the mark is gone. A pattern of the same logo repeated 40 times at 30 percent opacity, rotated 30 degrees over the busy parts of the image, cannot be cropped away without discarding most of the picture, and cannot be painted over without reconstructing the detail underneath. The math below explains why the second case is so much harder to defeat.

What a watermark actually changes

Every output pixel is a blend of two colors: the original photo pixel and the watermark pixel. The blend uses one number, the opacity \alpha, which runs from 0 (invisible) to 1 (fully opaque). The standard formula for the result color is:

Here C_{out} is the final color of that pixel, C_{mark} is the watermark color, C_{photo} is the original color, and \alpha is the opacity. The same formula runs once per color channel (red, green, blue). Where the watermark does not cover a pixel, \alpha = 0 and the photo passes through unchanged.

Work one channel. Say a photo pixel is red at value 200 and your watermark is white text at 255, with \alpha = 0.3. The result is 0.3 \times 255 + 0.7 \times 200 = 76.5 + 140 = 216.5, which rounds to 217. The pixel got lighter by 17 out of 255, about a 7 percent shift. That small, consistent shift is what your eye reads as a faint overlay.

When to use it, and when not to

Use a watermark when you publish images you want credited and you accept that the mark will be visible. Portfolio shots, proof galleries before payment, and social posts are the usual cases. A tiled, semi-transparent mark over image detail is the format that resists removal best.

Do not use a watermark to hide sensitive content. Opacity below 1 leaves the underlying pixels partly readable, and even an opaque mark can sometimes be reversed if it is uniform. To truly destroy information in a region, use the Blur / Pixelate Redactor, which replaces pixels instead of blending over them. And if you only want caption text top and bottom, the Meme Generator is the simpler fit.

Coverage: single mark versus tile

The protective strength of a watermark scales with how much of the image it touches. A single mark covers a fixed area. A tile repeats the mark on a grid, so coverage grows with the number of repeats.

If your mark is w by h pixels and you tile it on a grid with horizontal spacing s_x and vertical spacing s_y across an image W by H, the number of tiles is:

N is the tile count, and the brackets mean round up so the pattern reaches every edge. On a 2000 by 1333 image with a mark spaced every 300 pixels horizontally and 250 vertically, you get \lceil 2000/300 \rceil \times \lceil 1333/250 \rceil = 7 \times 6 = 42 copies. To crop out all 42, an attacker would have to cut the image down to almost nothing.

Worked example with the demo data

How opacity reads on different backgrounds

The same opacity looks different over dark and light areas. Over black (0), white text at 30 percent gives 0.3 \times 255 = 76.5, a clear gray. Over white (255) it gives 0.3 \times 255 + 0.7 \times 255 = 255, invisible. That is why a single-color watermark can vanish on a matching background. Placing the mark over detailed, mid-tone regions keeps it readable and, as a bonus, makes it harder to paint out.

Size and rotation

Size is set as a fraction of the image, so a mark scales with the photo instead of shrinking on large exports. If you set the mark to 25 percent of image width on a 2000 pixel wide image, the mark is 500 pixels wide. On a 1000 pixel export it becomes 250 pixels, keeping the same relative footprint.

Rotation matters for removal resistance. A horizontal mark aligns with typical crop and content-aware fill boundaries. A mark at 30 or 45 degrees cuts across those boundaries, so an eraser must follow a diagonal through varying detail. Rotation does not change the blend math, only which pixels the mark covers.

Common mistakes

Opacity too low to survive re-saving

At 5 percent the per-pixel shift is only 0.05 \times 255 \approx 13 at most. JPEG compression at quality 70 can move nearby pixels by a similar amount, blurring the mark into noise. Keep visible marks at 20 percent or higher.

Single corner mark on croppable content

If the subject sits in the center, a corner mark is free to crop. Center-crossing tiles remove that option.

Solid color over a matching background

White on a bright sky, black on shadow: the blend leaves almost no shift. Use a mark with an outline or a contrasting stroke, or tile so at least some copies land on mid-tones.

Marking after resizing

If you watermark first and downscale later, small text can become unreadable. Watermark at the final output size.

Related tools

To hide content rather than claim it, use the Blur / Pixelate Redactor. For top-and-bottom caption text, the Meme Generator is quicker. If your logo has a solid background you want gone before stamping it, run it through the Background Remover first so only the shape blends onto the photo.

Related tools

Screenshot Annotator

Crop, then add arrows, boxes, blur, highlights and text to screenshots.

Social Media Image Cropper

Crop to Instagram, X, Facebook, YouTube & LinkedIn sizes.

Background Remover

Remove solid or near-solid backgrounds to transparency.

Blur / Pixelate Redactor

Blur or pixelate regions to hide sensitive content.

Brightness & Contrast Editor

Fix dark or washed-out photos with simple sliders.

Image Noise Remover

Clean up grainy photos with median and smoothing filters.