Image descriptor quantization
AssigneeMicrosoft Corporation
InventorSimon Winder
Life of this patent
Estimating the term…
Abstract
Image descriptor quantization technique embodiments are presented which quantize an image descriptor defined by a vector of number elements. This is generally accomplished by lowering the number of bits per number element to a prescribed degree. The resulting quantized image descriptor exhibits minimal loss of matching reliability while at the same time reducing the amount of storage space needed to store the descriptor in a database. Lowering the number of bits per number element also allows for increased matching speed.
Claims5
A computer-implemented process for quantizing an image descriptor of a type comprising a vector of number elements, said process comprising using a computer to perform the following process actions:
- normalizing the image descriptor to a unit vector;
- clipping all the descriptor number elements that exceed a prescribed threshold to the value of the threshold to produce a clipped image descriptor;
- re-normalizing the clipped image descriptor to a unit vector;
- repeating the normalizing, clipping and re-normalizing actions a prescribed number of times;
- for each number element,
- dividing the number element by a prescribed quantizing factor to produce a resulting quotient, and
- replacing the number element with the next lower integer number to the resulting quotient;
- and
- designating a resulting vector as representing a quantized image descriptor.
Dependencies are inferred from the claim text.