Binary Descriptors

One of the issues with traditional descriptors is that they are composed of a vector containing a certain number of values, which are typically floating-point numbers due to possible normalizations. As a result, both the extraction of this vector and the subsequent comparison phase require significant computational time.

One of the most promising alternatives is to extract a binary vector as a descriptor. The binary vector occupies less memory space, and to perform the comparison, it is sufficient to calculate the Hamming distance between the respective binary strings. The Hamming distance is computed very efficiently by performing the XOR operation on the binary strings and counting the active bits (POPCOUNT).