ImageMagick: Command-Line Image Surgery

ImageMagick is a scriptable image-manipulation toolkit — think Photoshop operations you can run from the terminal and loop over. In CTFs it earns its keep for the unglamorous but constant tasks: stitching image fragments back together, cropping a region out of a larger picture, converting formats, and flattening layers. No GUI, no clicking, fully reproducible.

Two Brunner 2026 challenges used it directly: Magic or Not (stitching decoded strips into the final image) and Business Trip (cropping a region for closer inspection).

Installation

# Debian/Ubuntu
sudo apt install imagemagick
# macOS
brew install imagemagick

Modern ImageMagick (v7) uses a single magick command; v6 uses individual commands like convert, identify, montage. Both are shown below — if convert is missing, prefix with magick (e.g. magick convert ... or just magick ...).

Inspect first: identify

Before transforming anything, know what you have — dimensions, format, colour depth:

identify image.png
#   image.png PNG 800x600 800x600+0+0 8-bit sRGB 45KB
identify -verbose image.png    # everything: geometry, channels, embedded profiles

Stitching images together (append)

This is the Magic or Not move: you’ve decoded several horizontal strips and need them recombined into one image.

# horizontal — place images left to right
convert strip1.png strip2.png strip3.png +append combined.png

# vertical — stack top to bottom
convert row1.png row2.png row3.png -append combined.png

Mnemonic: +append joins side by side (columns), -append stacks (rows). Glob works too, but mind the ordering — shell globbing is lexical, so name fragments 00, 01, ... 10 (zero-padded) or list them explicitly:

convert $(ls -v strip*.png) +append combined.png

Cropping a region

The Business Trip move: pull a specific rectangle out of a larger image (a sign, a landmark, a corner with a leak).

convert input.jpg -crop 300x200+50+80 region.jpg
#                        │   │   │  └ y offset from top-left
#                        │   │   └── x offset from top-left
#                        │   └────── height
#                        └────────── width  (WxH+X+Y)

To slice an image into a grid of equal tiles (handy when a challenge splits a flag across cells):

convert input.png -crop 100x100 tile_%02d.png    # every 100x100 block -> tile_00.png ...

Format conversion and everyday transforms

convert image.webp image.png          # convert format by changing the extension
convert image.png -flatten out.png     # merge layers / remove alpha (reveals hidden layers)
convert image.png -negate inverted.png # invert colours — occasionally reveals hidden text
convert image.png -rotate 90 out.png
convert image.png -resize 200% big.png # upscale to read tiny detail
convert a.png b.png -compose difference -composite diff.png   # spot-the-difference

-flatten and -negate are quietly useful in stego-adjacent challenges: flattening can expose content hidden in a transparent or non-visible layer, and negating sometimes surfaces low-contrast text.

montage: contact sheets

When you have many fragments and want to eyeball them at once (order, orientation, which piece goes where):

montage tile_*.png -tile 4x -geometry +2+2 sheet.png

A word on the security policy

ImageMagick ships a policy.xml (often under /etc/ImageMagick-*/) that restricts risky coders and resources — a response to past vulnerabilities (e.g. “ImageTragick”, ghostscript-backed formats). If a legitimate operation is refused with a not authorized error, that policy is why. Loosen it only for files you trust, and prefer working in a throwaway directory or container when handling challenge-supplied images.

Best practices

• identify before you transform. Knowing exact dimensions makes crop/append offsets exact instead of trial-and-error.
• Zero-pad and sort fragment names. +appending globbed strips in the wrong order is the most common way stitching goes sideways — use ls -v or explicit ordering.
• Keep the original. Always write to a new filename; ImageMagick will happily overwrite in place.
• Reach for -flatten / -negate / difference when a “normal-looking” image is the whole challenge — the hidden content is often one transform away.

Conclusion

ImageMagick is the duct tape of image challenges: rarely the star, constantly needed. Whenever you’ve got pieces to reassemble, a region to isolate, or a format that won’t open, one convert line beats a round-trip through a graphics editor — and it’s repeatable when you inevitably get the offsets wrong the first time. Happy converting!