General Note: The detailed explanations remain in the parameter files. The following notes express the key points of each experiment in the pursuit for a goal. Goal 1: Getting decent results of NRR using res_info as a texture OF. 160102005-1 (on Baine) - Testing compilation and run of new tools. 260105-1 (256 bins, filter of size 5) 260105-2 (256 bins, filter of size 50) 260105-3 (256 bins, filter of size 100) 260105-4 (256 bins, filter of size 20) Results: In all the above, results are good so it appears SSD was used rather than res_info. Res_info was intended. Note: No proper matcher above? Maybe res_info is not used at any of the registration stages? 260105-5: Trying comp_region_matcher, not sequential_matcher Thoughts: It seems like the previous experiments towards the end, but because of the matcher type, warps are applied in reverse order. 260105-6: Trying res_info at all registration stages Results: The results obtained are now reminiscent of old results from Baine (see end of December 2004). This means that proper res_info can now be investigated. Note: Warp aggregation order needs to be reversed possibly. 260105-7: Same as the previous experiment, but pair-wise stages only, all with res_info (256 bins, filter of size 5). 260105-8: As above, but res_info (256 bins, filter of size 20). 260105-9: As above, but res_info (256 bins, filter of size 50). Results in 260105-7-9: Quite poor. The idea of changing filter width to improve results is hence given up. 30/01/2005 - 01/02/2005: Experiments in 2-D with data from Mike Rogers 020205-1: Checking to see if res_info is stable at the solution. This is done by registering a volume onto itself. Results: No warps are applied Conclusion: Minimum is at the solution (which yet cannot be reached properly).