Liquid Rescale plug-in updated for GIMP 2.8

Liquid Rescale plug-in updated for GIMP 2.8

Carlo Baldassi released an update of Liquid Rescale plug-in for GIMP that implements so called content-aware scaling.

In the unlikely case you are not familiar with the technique, the idea is to protect parts of an image from transformation while intelligently stretching or shrinking other parts. The outcome is an image with adjusted composition and very little, if any, visible glitches.

The only change in this release is the switch to a new object API in GIMP 2.8. That should remove incompatibility warnings for users of v2.8. An earlier version that's compatible with v2.6 is still there to download. In fact, the newly released v0.7.2 isn't available as an installer for Windows yet.

Now for the interesting bit. Carlo has been lacking the spare time to work on Liquid Rescaling for a while now and thus cannot implement any of ideas that Peter Sikking's usability students came up with last year.

So the plan is to move Liquid Rescale to the upstream GIMP project and make it part of future releases (beginning with 2.10). This should stimulate its further development. The initiative has been approved by the GIMP team.

The first stage would be porting Liquid Rescale to GEGL. The backend of the plug-in, a library called liblqr, is already quite capable in terms of handling high dynamic range images and supports up to 64bit per color channel (GEGL only does 32bit float/integer per component as of today).

If you are interested to work on Liquid Rescale as an integral part of GIMP, get in touch with either Carlo, the GIMP team, or better yet, both.

Was it useful? There's more:

2 Responses. Comments closed for this entry.

  1. Second that spambot :D

  2. Gegl(Buffer) can store 64bit floats, just use the Babl type “RGBA double”*.
    However, most existing GEGL operations operate on 32bit floats.

    * Or any other datatype really, as Babl extensions can be registered on the fly. One is required to provide a conversion from/to 32bit float - and conversions to any other format will just work, albeit possibly a bit slow.