With more time, I would develop further these projects (if it inspires somebody, please let me know):
Real-time interactive physics simulation for teachers
The direct visualization of physics experiments, and the ability to play with them, increase the understanding of physics phenomenons. Covered areas could include mechanics (see Phun), optics, electromagnetism, waves, etc. In my opinion, there is a big potential for education in the simulation of the wave equation: phenomenons such as optical diffraction, meta-materials, and in general the propagation of waves in any non-linear medium could be played with.
Podcatcher with local web-page processing
Civilization-like game with evolution of living cells
The player would be able to control a growing creature, and evolutions would increase the available functionality (e.g. photo-sensible cells would become eyes).
Picture-by-picture comics reader
On small displays, reading a comic book is difficult because the zooming level must be high to be able to read the text, and the viewing window is hence small. A program could analyze each page on-the-fly and zoom on each image automatically (meanwhile there is a similar application on iPad, but the comic pages must be prepared).
Indoor auto-mapping per video
The application would propose to capture a video along with the recording of the position of the device (e.g. a mobile phone). After processing, a floor map of the visited rooms would be displayed. The idea here would be to use the motion compensation computation hardware (available in any modern video recording device producing mp4) in order to estimate picture changes due to parallax.
Stock indicators with simulation
Working with rapidly changing financial products is a full-time job, and the use of simple alerts (e.g. based on thresholds) is not a practicable strategy. Another approach would involve the design of programmable buy/sell indicators that each user could create on line. Indicators could be tested on the current stock values before being used for real operations. The ultimate goal would be to grow a community around these indicators, allowing users to compare indicators, bring attention to the best ones (top 10 charts) and exchange information. The variables used by indicators should include the widest range of information (stock values, news, expert opinions, Fourier transform of historical values, book of order, etc).
Information system design tool
Model-based engineering methods based on UML or SysML focus often on the visual documentation of a project. The different diagrams can be created independently, and depict as many aspects of the same system, but without a common model. As a result, inconsistencies cannot be detected automatically, and the system cannot be validated or simulated. A model-based engineering tool should focus on having a running model, from which diagrams and requirements should be exported (and not vice-versa).
Virtual origami designer
The practice of origami (paper folding) is a nice occupation but diagrams are difficult to decode and learn. A tool could help the user to see an animation of given diagram steps (with zoom and rotate capabilities), and support design functions that would allow to start at any step of any existing diagram.
I feel I shouldn’t give away all these ideas but to be realistic there is a little chance that I ever attempt to realize one of them, so it is better to have them on-line for anybody interested. And well, there is nothing ground breaking here.