COMIS projectPart of the project is the downloading of the decoder software and the real transit of the video bitstream over ISDN.
![]() Figure 1 |
Figure 1 demonstrates how an image can be described using a binary segmentation tree, while figure 2 shows the result of the algorithm on an orignal image(left), with the reconstructed image (center) and the associated multigrid (right).
![]() Figure 2 |
The inter-images are mainly encoded by a reliable motion field analysis and transcribed in a way similar to the intra transcription. The algorithm received thus the name of COMIS (Coding On Multigrids Image Sequences) and was essentially a spatio-temporal block-based coding technique.
Indeed the only scalabilty that was possible with the first version of COMIS is depicted on figure 3 (left: original, center: coarse multigrid, right: first step towards a new scheme).
![]() Figure 3 |
A reorganization of the scheme towards an improved coding efficiency and object-oriented functionnalities has thus been performed, as depicted on figure 4: a region model box is present in both the encoder and the decoder sides, and allow the scheme to deal with objects or regions as abstract entities.
![]() Figure 4 |
![]() Figure 5 |
![]() Figure 6 |
For more information on our researches in Image Analysis and Compression, please contact Benoît Macq
Page: UCL | FSA | ELEC | TELE | Useful links