Projects / Motiontrack

Motiontrack

Motiontrack is a set of tools that detects motion between two images. It is able to successfully distinguish random flicker from real object movement by applying a set of filters to the images, and can optionally ignore given colors and/or image regions. The roadmap provides for being able to identify objects by pattern detection and AI routines. Currently, this part is able to turn line-art images into ASCII-art text and auto-align images which are slightly staggered or contorted. This might also be useful to correct the perspective of image sources from different 3D angles.

Tags
Licenses
Operating Systems
Implementation

Recent releases

  •  28 Nov 2004 03:34

    Release Notes: This release now supports multi-processing when compiled with appropriate flags. It then forks into several processes, each computing an equal part of each image manipulation (for example filtering) task, and then synchronises the result to the parent process via half duplex Unix pipes. In this way it possibly improves performance on SMP CPUs, multi-CPU systems, or OpenMosix-like cluster solutions.

    •  20 Oct 2004 13:04

      Release Notes: Movement Sectors are now visually outlined in the output image, and the speed of the sector check has been improved. Note that the names of most executables were changed to more unique ones to prevent conflicts with other projects, so scripts using motiontrack might have to be altered.

      •  17 Oct 2004 14:41

        Release Notes: Motiontrack can now auto-normalize and auto-align to be compared images. This includes small rotations and/or perspective corrections of low angle. It's also possible to tweak the sector difference detection algorithm to work better for auto-comparison of generic images.

        •  21 Mar 2004 14:38

          Release Notes: A complex re-design of the image manipulation filters has been done, supporting fuzzy tune values for almost everything now. Image resizing can now be done in any percentage of the original size. A brush size can be specified for the "enhance" and "blur" steps, tuning their behavior. "enhance" now works more like a real image frequency derivation, and additional values can be set to optimize speed performance. As always, some bugfixes have been made.

          •  19 Mar 2004 10:07

            Release Notes: The ASCII art pattern detection now uses a real ASCII font with all available characters. Additional filter programs and steps were added to change image sizes, and several bugs were fixed and optimizations made.

            Screenshot

            Project Spotlight

            OpenStack4j

            A Fluent OpenStack client API for Java.

            Screenshot

            Project Spotlight

            TurnKey TWiki Appliance

            A TWiki appliance that is easy to use and lightweight.