SIP x SIVP x IPD
Ricardo Fabbri
rfabbri at gmail.com
Fri Apr 13 08:28:50 CEST 2012
Hi,
Scilab has 3 major toolboxes for image processing. I quickly compiled a
list which tries to describe what is special about each of them, focusing
on SIP.
Let me know if you think any item should be added or changed. I put this
list on SIP's home page as well, and encourage IPD and SIVP developers to
display a similar list on their project pages as well.
*SIP <http://siptoolbox.sf.net/> vs. SIVP <http://sivp.sf.net/> vs.
IPD<http://forge.scilab.org/index.php/p/IPD>- the 'other' Scilab Image
Processing toolboxes
*
Basically, SIP aims towards comprehensive functionality. The only price is
a longer installation process to get all the features due to many
third-party dependencies.
- SIP prioritizes GNU/Linux. SIVP and IPD are *currently* easier for
Windows users, followed by SIVP.
- SIP provides unified bindings to several image processing libraries:
ImageMagick, OpenCV, animal, and (soon) Leptonica.
- SIP is targeted to more advanced users: aims at more functionality
than SIVP and IPD, but can be harder to install. This is a design decision.
- SIP has the most number of functions. As of april 2012, SIP has 74
help pages, compared to 55 from SIVP and 53 from IPD.
- SIP appeared first, SIVP as a friendly fork of SIP. IPD appeared more
recently. Most of the SIVP improvements have recently been merged back into
SIP.
- SIP is designed for very rapid prototyping of imaging solutions.
- SIP has a friendly and responsive developer and user community.
- SIP provides ample illustrated documentation with examples.
- SIP provides state-of-the-art Euclidean morphology-related algorithms,
such as dilations, erosions, distance transforms, skeletons, watershed
segmentation, with reference implementations that are mostly superior to
Matlab and other software.
- SIVP has explicit handling of integer pixel depths, while SIP is
purposedly built for double representation, for simplicity.
- SIVP *currently* has support for video processing beyond SIP.
- SIP focuses on ease of use, functionality, and on the *internal* speed
of the provided functions, *not* on any low-level user-visible details
that could complicate quick usage, such as specialized support for huge
images or integer pixel depths. Most image processsing production code or
complex time-consuming algorithms that need to deal with very large images
should switch from Scilab to C/C++ (for example) as soon as the idea is
working for medium-sized images in Scilab+SIP.
- SIP may be part of Google Summer of Code 2012!
- *Conclusion: All 3 Scilab IP toolboxes are great and complement each
other. They even share code. *
Best regards,
Ricardo Fabbri
--
Linux registered user #175401
www.lems.brown.edu/~rfabbri
pt.wikipedia.org/wiki/IPRJ
labmacambira.sf.net
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.scilab.org/pipermail/users/attachments/20120413/e197e986/attachment.htm>
More information about the users
mailing list