Revision history [back]

I realize that FREAK uses a (512 bit?) binary descriptor. What I dont understand is the relevance of that format for the purposes of matching. Is it like BREIF where each bit is independent? Or, is it more like the original paper suggests, with the values representing coarse to fine grain comparisons? If so what is the bit ordering and chunk size?

The demo code for OpenCV using the standard Brute Force Hamming Distance matcher which will work for any binary descriptor but the original paper suggesting using a saccadic search. I don't believe such a search is built-in to OpenCV.

For my particular application, being able to do a saccardic search would be ideal. I can certainly code one but I need to understand the format of the descriptor first. I am investigating the freak.cpp code in OpenCV but as with most optimized complex algorithms, sorting out the guts is taking a while.

I realize that FREAK uses a (512 bit?) binary descriptor. What I dont don't understand is the relevance of that format for the purposes of matching. Is it like BREIF where each bit is independent? Or, is it more like the original paper suggests, with the values representing coarse to fine grain comparisons? If so what is the bit ordering and chunk size?

The demo code for OpenCV using the standard Brute Force Hamming Distance matcher which will work for any binary descriptor but the original paper suggesting using a saccadic search. I don't believe such a search is built-in to OpenCV.

For my particular application, being able to do a saccardic search would be ideal. I can certainly code one but I need to understand the format of the descriptor first. I am investigating the freak.cpp code in OpenCV but as with most optimized complex algorithms, sorting out the guts is taking a while.

I realize that FREAK uses a (512 bit?) binary descriptor. What I don't understand is the relevance of that format for the purposes of matching. Is it like BREIF where each bit is independent? Or, is it more like the original paper suggests, with the values representing coarse to fine grain comparisons? If so what is the bit ordering and chunk size?

The demo code for OpenCV using eses the standard Brute Force Hamming Distance matcher matcher, which will work for any binary descriptor descriptor, but the original paper suggesting suggests using a saccadic search. I don't believe such a search is built-in to OpenCV.

For my particular application, being able to do a saccardic search would be ideal. I can certainly code one but I need to understand the format of the descriptor first. I am investigating the freak.cpp code in OpenCV but as with most optimized complex algorithms, sorting out the guts is taking a while.

I realize that FREAK uses a (512 bit?) binary descriptor. What I don't understand is the relevance of that format for the purposes of matching. Is it like BREIF where each bit is independent? Or, is it more like the original paper suggests, with the values representing coarse to fine grain comparisons? If so what is the bit ordering and chunk size?

The demo code for OpenCV eses ueses the standard Brute Force Hamming Distance matcher, which will work for any binary descriptor, but the original paper suggests using a saccadic search. I don't believe such a search is built-in to OpenCV.

For my particular application, being able to do a saccardic search would be ideal. I can certainly code one but I need to understand the format of the descriptor first. I am investigating the freak.cpp code in OpenCV but as with most optimized complex algorithms, sorting out the guts is taking a while.