surfCon.TODO

TODO
----

UI
--

- no longer need the vertices and faces in the ds file, i think.

- add greyscale option

- add transparency on the movable slide placed on top of the target?

- view contour points in 3d (rotation and that stuff), click on points to go to slice for contour edit

- undo

- add in a little more "friction" in the tack placement, i.e., slow
  down the mouse movements?

- log files

- basis functions for interpolation between contour points

- accelerated libs - os stuff

MESHING
-------

- eliminate dependence on triangle by capping tubes by adding a fuchs
  generated surface between the last contour and a vertex added at the
  centroid of the contour with a z coordinate midway between
  slices. 

- when meshing open to closed contours, or peices of closed contours
  to other closed contours, or capping contours, allow direction of
  tiling on the open contour to reverse at the end of the
  contour. this will allow linear contours to tile to loops if they
  open up. similarly, allow this to happen in normal circumstances
  when tiling two open contours so that tiling of the ends of contours 
  can be better.

- when handling branches, if there is a postbranch contour for which
  no vertices in the prebranch contour are closest, segfault will happen
  due to degenerate transition condition

- problem where closed contours that have been split are treated as
  open and sometimes cause a weird twisting of the surface

- check out http://www.cs.duke.edu/~edels/TriTop/

- make the datastructures for fuchs compatible with the surf geom toolbox
  and just better overall

- implement reeb graph construction for correspondence problem
partially done (facility for manual identification)


branching ideas
---------------

* come up with some sort of transform (fourier?) that represents the
  shape of a contour. use this shape transform to determine
  probability that each contour is connected. use thresholding to
  determine existence of branches and to connect correct segments. use
  exhaustive search of contour spliting hypotheses when below
  threshold. shouldn't be too much of a performance hit if there are
  few branches compared to the number of slices. 

* combined solution to the correspondence and branching problems:
1) sort contours based on similarity metric to contours on next slice
2) pair contours in decending order of similarity, remove both
   contours from further pairing.
3) below some threshold (fix this idea: thresholds are bad) decide
   that contours are too dissimilar to be paired. 
4) parcelate each contour into regions of similarity to other contours
   and their regions. this will be the hard part. it could be very
   time consuming to search all possible regions on both contours and
   compare them. 
5) sort contour segments by similarity and pair as before
6) tile
7) close holes

* when a closed contour is a post branch contour, find transition by
  closest approach instead of my misclassifications. right now the
  misclassification method is producing face intersections since the
  prebranch contour may have vertices after the closeest approach that
  curve around and should be part of the adjacent open contour. see
  problematic/closed_post_branch.obj


DONE
----

- implement simple solution for branching problem, at branches, just connect
  nearest neighbors 
done 2005-03-18 (see notes for implementation details)

- dynamically load textures to get around memroy problems
done 2004-11-01

- blank (no textures when too much memory is used)
not fixed, but greatly decreased (factor of 10) the memory usage 2004-07-23 
done 2004-11-01

- offset should remain across images in tack mode
done 2004-07-23

- implement contour breaking
done 2004-07-23

- implement dragging mouse places contours with timing or spacing
done 2004-07-23

- crash after load of 368 stria slices (512x512), 
  seemed to be plenty of memory available. actually ran out of memory.
done 2004-07-23

- fix the tack selection
done 2004-07-23

- decrease sensitivity of zoom
done 2004-07-23

- window resize seems to make app VERY sluggish
done 2004-07-22

- crash if tack mode entered immediately
done 2004-07-22

- tack size should remain constant regardless of scale
done 2004-07-22

- users specifies list of files instead of renaming and copying all images
done 2004-07-21

- can you make the line segments terminate outside of the marker hole 
  such that you don't see the line intersection? (actually, maybe this is 
  a feature and not a bug....)
done 2004-07-21

- the window scrolling (or panning) is not only controlled by the number 
   pad keys, but also the standard numeric keys along the top of the 
   keyboard. i think that there is a way to differentiate so that the 
   top-row number keys will not cause the frame to pan.
determined impossible 2004-07-21

 - how do you view the tack contours on adjacent slices? also, maybe you 
   should allow the user to see the contours when in align mode if they 
   want to tweak the alignment after carefully tacking out features. so, 
   for example, if i meticulously tack out layer 4 in two adjacent slides,
   i could adjust the alignment to ensure that the tacks are aligned,
   instead of the tissue boundaries which will probably work out to be the
   most accurately aligned portion of the slice when using eric's
   'flicker' method on the images alone.
done 2004-07-21

 - so it looks like you fixed the $Id: surfCon.TODO,v 1.27 2005/10/28 23:18:28 oph Exp $ issue by just deleting the first 
   character. the trailing dollar sign is still there, and the string is
   less recognizable as a CVS keyword expansion. you know, you can turn
   off keyword expansion in CVS by adding the file with the '-kb' flag, as
   you would if the file were binary. that could be dicey if the user ever
   wants to check in a 'ds' file. maybe you could do a 'Id' -> 'ID'
   replacement, since the keyword substitution is case sensitive.
done 2004-07-21

 - there are some misspellings in the README. in particular, you spell 
   'movable' as 'moveable', which is think is valid but non-standard.
done 2004-07-21

 - even though there isn't much to the installation process, people 
   usually expect an INSTALL file, and a reference to the INSTALL file 
   from the README.
done 2004-07-21

 * i think what the README needs is a step-through tutorial and some more 
   verbose explanations of what the program does.
done 2004-07-21

 - the most important two function keys, 'A' and 'T' were hard to find in 
   the README. i was surprised when i first stated the demo and saw a 
   blank screen. it took me a minute to find out to press the 'A' key. 
   yes, i should RTFM, but there should also be a way to let the user 
   start up right away then consult the README as needed. (this is my 
   anticipation of the schwartzian impatience principle coming into effect 
   when eric runs this the first time.)
done 2004-07-21

 - i would take out the './' before 'surfCon' in the README since your 
   installation instructions say to copy the executable to somewhere in 
   the path, so it shouldn't be in the cwd. also, you might want to 
   suggest a standard install location, such as '/usr/local/bin'.
done 2004-07-21

 - when you introduce the 'dataset', its not clear that this is just a
   name that you use to describe the organizational unit consisting of a
   given set of images and the 'ds' file. the README might want to explain
   what a 'dataset' is and the format of the 'ds' file (including what 
   stuff it contains).
done 2004-07-21

 - the explanation of the '-d' flag in the README looks like its missing a 
   word.
done 2004-07-21

 - so the 'dataset file' is always located in the 'data directory'? i am
   confused about your command line example for creating/initializing a
   dataset.
done 2004-07-21

 - this might sound silly, but you might want to loose all those '#'
   characters that are all over the README. check out the NVIDIA 
   readme, especially the opening header and how it gives the version
   number, an overview of the README, etc. i think that the NVIDIA readme  
   is one of the easiest to read (although you won't need all the 
   appendices, but i like how it looks sparse and is easy to navigate). 
   make use of whitespace to format.
done 2004-07-21

 - right now you have Cntl-LEFTCLICK for delete and MIDDLECLICK for
   selecting endpoints of a segment. i found myself deleting when i meant 
   to be selecting. these keys seem like they want to be reversed.
done 2004-07-21

 - so i understand how you are cycling through tack marker and line
   segment color combinations to keep them all distinct, but it does look 
   kind of wacky.  :)  do you think that for the first cycle at least the
   tack markers and line segments could have the same color?
done 2004-07-21

 - any fine rotation?
done 2004-07-21

 - there's a typo in the menu for the 'r' key.
done 2004-07-21

 - i think that the thickness of the tack marker ring could be reduced. 
   thing were getting cramped when i tried tacking layers 1, 4, and 6.
done 2004-07-19

 - when you select a tack marker, a bulky blue ring surrounds the marker
   signifying that its selected. again, this made it difficult to place 
   the marker precisely. do you think that, instead of the surrounding 
   blue ring you could, say, make the selected marker white? the blue 
   surrounding ring was especially weird when the markers themselves were 
   blue.
done 2004-07-19

 - the key for placing a 'marker' is given in the README, and i assume 
   that these are for things like electrode lesions, but this is not 
   explained in the README.
done 2004-07-19

- scaling is broken for contours
done 2004-07-16

- add in some online help, i.e., display keys and commands in menu
done 2004-07-16

- make README more readable, especially expand upon the role of the
  "dataset" file, and make the options easier to find. watch terminology: 
  points, nodes, vertices, positions
done 2004-07-16

- view contours from adjacent slice(s) using alpha or something (preserve color) allow editing of points only on current slice
done 2004-07-13

- tell user that NumLock must be on for fine-scale translations in
  alignment
done 2004-06-28

- include how to exit in the readme. upon exiting, prompt the user to
  save the dataset (yes/no/cancel)
done 2004-06-30

- how to go to next/prev slide in alignment
done 2004-06-28

- reformat entries in README so that its obvious which key corresponds
  to which command
done 2004-06-28

- if 'n' provides an incorrect number of images, prog loads as many as
  it can, then crashes
done 2004-06-30

- big bad crash. after starting, i hit 'A'.
done 2004-06-30

[jonnyreb@alaya surfCon-0.1]$ surfCon -c -f dhtest/dhtest.ds
loading 0 slices images ...
GLUT: Warning in surfCon: glutReshapeWindow: non-positive width or
height not allowed
X Error of failed request:  BadValue (integer parameter out of range for
operation)
  Major opcode of failed request:  12 (X_ConfigureWindow)
  Value in failed request:  0x0
  Serial number of failed request:  55
  Current serial number in output stream:  55


- what are the default parameters for '-n' and '-s'?
done 2004-06-30

- add some "status" messages somewhere in window or command line, like
  "switching to taking mode" or "line segment selected. click on any
  position on segment to add in contour vertex".
done 2004-06-28

- vertex and segment colors
done 2004-07-08

- make vertex markers larger (annuli?)
done 2004-07-07

- allow switching of contours being edited
done 2004-07-08

- windows thing
done with cygwin 2004-06-28

- batch mode
done 2004-06-28

- config file
done 2004-06-28

- support multiple contours on same slice, and some sort of labeling scheme
done 2004-07-08

- fix $Id: surfCon.TODO,v 1.27 2005/10/28 23:18:28 oph Exp $ strings in "dataset" files
done 2004-06-28

- allow user to STOP flickering. its hard to know how to rotate when
  its not clear WHICH slide is being rotated and which is fixed. it
  would be cool to stop the flicker and allow the user to rotate on
  the still image. then swap the 'moveable' slide for the 'target'
  slide.
done 2004-06-28

- matlab thing to read in points, connect lines
done 2004-06-25

- adjust contrast/brightness/colormap smoothly in window to aid
  tacking/aligning?
done 2005-06-20

- don't write out empty align actions or tack lists
done 2005-06-20

- make it easier to find and fix self-intersecting contours
done 2005-06-21

- bug where app starts using all of cpu and being really
  sluggish after a few seconds of use. 
done 2005-06-21