pinpoint-team team mailing list archive

[Andrew Straw <strawman@xxxxxxxxxx>]

Hi,

I've just pushed an early version of a distortion-estimation GUI tool to
 lp:~astraw/pinpoint/dev . The program is run from the command line
after installing the pinpoint package by running "python
distortion_gui.py filename.jpg".

Implemented Features
--------------------

 * Simultaneously view multiple images sharing the same distortion model

 * Manually adjust distortion model and watch live updates of
undistorted image

 * Automatically estimate distortion using manually entered straight
lines. (Uses the algorithm of 'Line-Based Correction of Radial Lens
Distortion' (GMIP 1997) by Prescott and McLean)

Planned and wished-for features
--------------------------------

 * Automatically detected checkerboard corners so that distortion
estimation is fully automated.

 * Distortion visualization

 * An "add image" button - currently this must be done through the
command line at program invocation.

 * Alternative distortion models

 * A view of the lines that have been manually or automatically entered
on the image so that individual lines or vertices can be removed or
added. (Currently, there are a couple hotkeys in the image canvas for
implementing this.)

Known Bugs
----------

 * Images are shown in an expandable panel which is initially collapsed.
To un-collapse the panel, click along the top few pixels of the main
screen portion.

Thanks to Stéfan for providing a demonstration of how to use
scipy.ndimage.map_coordinates() for fast image warping. Also, this was
my first endeavor using enthought.traits, which is a really powerful
toolkit.