图片打马赛克

The image is cut into a grid of squares that starts at the top-left corner of your box, not of the photo, so the first block sits exactly on the edge you drew. Each square gets the mean red, green and blue of the pixels in it, weighted by their opacity so transparent parts of a PNG do not darken the result. Squares cut off by the box edge average only the pixels they contain. Nothing outside the box is read or changed.

With 64 px blocks, a 1200 × 800 photo is reduced to 19 × 13 squares, the last row and column only partly filled. At 8 px the same photo keeps 150 × 100 squares, enough to recognise a familiar face from across the room.

Pixelation is not encryption. Each block still holds the average of what was under it, and for text set in a known font that can be enough to work backwards: the open-source tool Depix (2020) recovered passwords from pixelated screenshots by matching block patterns against rendered text. The smaller the block compared with the letters, the more of the shape survives.

For anything you would not want read, such as a number plate, a card number, a street name or a document reference, switch Inside the areas to Solid black box. It paints the box opaque black and keeps no trace of what was there. If you prefer the pixel look for a plate, make each block larger than the height of a character, so no letter is split into more than one row of squares.

Every photo in the list keeps its own areas, so a crowd shot can have six boxes and the next picture just one. Select a file in the list to mark it. If a file has no areas while Selected areas is on, Apply skips it and labels it, instead of saving an unchanged copy you might post by mistake. Without a mouse, Add box places a box in the centre; arrow keys move it, Shift with the arrows resizes it and Delete removes it.

Whole image mode ignores the boxes and pixelates the entire frame, which is useful for a placeholder thumbnail or a deliberately blocky graphic.

Output can stay in the original format or switch to PNG, JPG or WebP; BMP files come out as PNG, and JPG and WebP follow a quality slider set to 92% by default. Flat blocks compress well, so a fully pixelated JPG is much smaller than the photo it came from: our 1200 × 800 test photo of 224,092 bytes, pixelated with the default 16 px blocks and saved at quality 92, came out at 11,026 bytes.

The new file is drawn from the decoded pixels, so EXIF data is not copied. Camera model, capture time and GPS position are left behind.