< ^ >

Implementation

Aim: build a quick/easy/light implementation of the above approach.

  • make it stand-alone from QBIC, but use QBIC's distance function

  • use the Unix dbm libraries to implement keyed retrieval

Two components:

build

  • reads image idents and color histograms from stdin

  • stores each in the Db, a dbm B-tree keyed on ident

  • computes m curve-points for each, and stores in Index, a dbm B-tree keyed on curveId and position (Ci(v) value)

query

  • reads a single color histogram from stdin

  • uses above method to find k best matches

  • display list of best matches idents and distances on stdout, along with performance statistics


< ^ >