ShrinkShot

A simple CLI tool to detect and cut out empty areas of a PNG image (typically: a screenshot) in order to make it smaller without resizing its content.

Official release: crates.io.

Features

Basic

The program simply scans for whole pixel columns and lines which are identical to their neigbours. These columns and lines can be cut out from the image, no information will be lost, but the image will be smaller and more compact.

Avoid background border

If you make a screenshot by marking the cut region by hand (sometimes called snipping tool), empty area detection may fail on the surrounding background. Shrinkshot uses a simple solution: just simply ignores 2-2 pixels at the borders (see IGNORED_MARGIN_SIZE).

Avoid cutting small areas

Texts are padded with empty pixel lines. Cutting these identical lines reduces the padding to 1 pixel height, which makes the text ugly and too dense. Vertical letter spacing can be ruined even more. To avoid this issue, the program does not cut small areas below a certain size (see MINIMUM_SHRINK_SIZE).

Avoid invisible gradients

Gradient and transparent window captions are empty areas with different pixels. To detect these areas, the program checks for similar colors, not for exactly the same (see PIXEL_CHANNEL_DIFF_SUM_THRESHOLD).

Installation

The V1 program required to install ImageMagick and pull upng submodule.

You can install it from crates.io:

cargo install shrinkshot

You can also build the app from source with cargo build, or let build.sh to do it.

Usage

Shrink specified file:

$ shrinkshot screenshot.png result.png

Pick last created file from a directory and add insert .shrnkd to its name:

$ shrinkshot -d ~/Screenshots

If any problem occurs, shrinkshot prints error messages to stderr.

You may assign a hotkey to yout script, which makes a screenshot of a window, passing through it on shrinkshot, and saves it.

Known issues

• The neighbour detection is somewhat naive.
• The program often makes text too dense.

Algorithm enhancements

It would be great to split the image to more regions (e.g. horizontal stripes), and cut out (same width) areas from it at different (horizontal) positions.

The most known use case of it is the status bar, which should be cut separately:

This image has no columns to cut (only lines):

  ---------------------------
 | content           content |
 | on left           on the  |
 |                right side |
 |                           |
 |                           |
 |          status           |
  ---------------------------

But it should be:

  ----------------------
 | content     content |
 | on left     on the  |
 |          right side |
 |                     |
 |                     |
 |       status        |
  ---------------------

The "status" line occupies small vertical size, so it's a good candidate to handle differently from the main area.

Empty areas marked with numbers (vertical only):

  ---------------------------   --
 | content 111111    content |  |
 | on left 111111    on the  |  | large area
 |         111111 right side |  |
 |         111111            |  |
 |         111111            |  --
 | 22222222 status 222222222 |  | small area
  ---------------------------   --

The program should detect main area (primary target for shrinking), and try to shrink same amount at smaller areas.

Credits

The idea and some sample images come from a question issued by @Thomas on the Software Recommendation (StackExchange) site.

Copyright

Use it as you want. I'll be happy if you integrate it and mention it in the credits.