Experiments 24112004-25-32

Checking different shape OF's, texture remains abs_diff

///////////////////////////////////////////////////////////

// 24/11/2004: Trying to see how shape OF's behave

// Comparison with various other methods

// 20 intensive iterations of NRR, two levels

// of affine.

1. 24112004-25: zero
2. 24112004-26: ssd
3. 24112004-27: abs_diff
4. 24112004-28: wtd_abs_diff
5. 24112004-29: eigen
6. 24112004-30: def_energy
7. 24112004-31: wtd_def_energy (NOT IMPLEMENTED?)
8. 24112004-32: shape

Conclusion: See individual results in the directory which collects RESULTS from all; Quality depends on type of analysis.

zero,ssd and abs_diff give different yet apparently equally good results. something goes quite wrong with wtd_abs_diff because there appears to be a very fuzzy results where the images are unaligned. eigen causes the image to disappear, def_energy gives a very good answer if not the best. No results have yet been returned for shape. it is very time-consuming.

For all results in brief see: SHAPE_OFs-24112004 in experiments archive where all final results have been put.