Remote Sensing Tools

There’s been a great deal of buzz for a couple months, concerning a technology known as Seam Carving — also being refered to as Content Aware Image Resizing or Aware Image Resizing. I’ve been experimenting with this for awhile, and I’m seeing some wonderful potential and practicaliy in its use — though, at the moment, in a limited way relative to satellite/aerial imagery. The methodology was developed and introduced to SIGGRAPH 2007 by Dr. Shai Avidan of Mitsubishi Electric Research Labs (MERL), and Dr. Ariel Shamir of The Interdisciplinary Center, Herzliya & MERL.

The implications of use are quite obvious to those who have experience in creating large mosaics of imagery, specifically in Remote Sensing & GIS. I can see various value-added potentials to the approach being used — from more accurate mosaic blending methods, to accurately replacing pixels that are desired to be excluded (objects undesirable to final product imagery, such as cars, boats, machinery, etc). In essence — being able to obtain, for lack of a better term, ‘bare Earth’ or ‘pure image’ results taking away image-object disruption. One can conclude as well that privacy issues could be relaxed, such as the case with Google StreetView imagery — where there would be a benefit to cutting out objects, such as persons or vehicles. To the extent that this is possible with different sources of imagery is more complicated — but this approach could be a reasonable starting-point.

The idea that I’m bouncing around currently, is actually more relative to atmospheric dehazing or ‘correction’ of the imagery. One of the algorithms demonstrated in the YouTube video automatically jumped-out as something I was working on similarly last year. With the added concept of vertical and horizontal ’seam carving’, however, it makes the image even more dynamic in being able to differentiate within the intensity maps — how-much and where.

Below is a current round-up of what I’ve researched happening since the paper was released. And as a note — Dr. Shai Avidan was hired by Adobe not long after this methodology was released, thus it’s probable that we’ll see the technology incorporated into their family of image editing products (and perhaps likely in a highly optimized manner that requires fewer resources).

The paper and a link to videos are made available at the following:

Dr. Ariel Shamir’s Web site:

http://www.faculty.idc.ac.il/arik/

See:

Shai Avidan, Ariel Shamir
Seam Carving for Content-Aware Image Resizing
ACM Transactions on Graphics, Volume 26, Number 3,
SIGGRAPH 2007

A link to both the video and the paper, conveniently at its own domain.:
http://www.seamcarving.com/

Current Projects Implementing Seam Carving

Standalone Programs

ImageSeams (Java 1.6 Webstart, Java 1.6 Swing GUI, Java 1.6 Swing GUI with Windows launcher):
http://www.semanticmetadata.net/ (Link shortened for display.)

EIkke Blog (Python)
http://blog.eikke.com/ (Link shortened for display.)

Fast Content Aware Image Resizing (OCaml):
http://eigenclass.org/hiki/seam-carving-in-ocaml

Liquid Resize (Standalone Prototype — Windows, Linux)
http://www.thegedanken.com/retarget/

Content-Aware Image Resize Project, Alex Kennberg (Windows Binary):
http://www.kennberg.com/project_resize.php

Tim Wintle’s Seam Resizer:
http://www.timwintle.co.uk/apps/seamresizer/

Plug-ins

Picutel Smart Resize (Photoshop, Windows [Mac version in 2008]):
http://www.picutel.com/
(Limited Size Free, Unlimited Commercial $98)

Liquid Rescale (GIMP):
http://liquidrescale.wikidot.com/
(Reported to be highly buggy.)

Flash/Actionscript, Web

Joe Ebert’s source (Actionscript):
http://blog.je2050.de/2007/09/02/content-aware-image-resizing/

Mario Klingemann’s optimized version (Actionscript):
http://www.quasimondo.com/archives/000652.php

Woodpecker (Joe and Mario’s collaborative project to create a web-based tool using the Aviary framework and Flex.):
http://www.creationonthefly.com/blog/20

Patrick Swieskowski (Web-based Seam Carver Demo):
http://swieskowski.net/carve/

Rsizr.com (Free processing service Web application):
http://rsizr.com/