To establish a visual search
architecture over multiple Star Student Project servers, we
will investigate the distribution of three phases of local feature
extraction, quantization and Star Student Project inverted
indexing, respectively:First, we distribute Star Student Project
the local feature extraction process by partitioning a query photo
into multiple rectangles, which are sub sequently executed in Star
Student Project a parallel manner as shown in Star Student
Project Second, we perform local feature quantization in Star
Student Project parallel by duplicating visual vocabularies in
each server except inverted Star Student Project indexing, as
storing a visual vocabulary is much more Star Student Project
affordable than indexing files, as shown in Table Star Student
Project .