DEVICE AND METHOD FOR EXTRACTING FEATURE POINT INFORMATION OF LARGE-CAPACITY SOURCE IMAGE BY USING SIFT ALGORITHM

摘要

PROBLEM TO BE SOLVED: To perform a SIFT (scale-invariant feature transform) algorithm by allowing a video having an arbitrary resolution to be processed by using an SRAM and DRAM in parallel without significantly increasing SRAM capacity.SOLUTION: A feature point information extraction device 100 includes: a first interface unit that reads respective parts of a first source image stored in a large-capacity temporary storage device, and generates at least one or more tile images from the first source image; and a feature point information extraction unit to which the at least one or more generated tile images are input, and which applies feature transformation to the at least one or more input tile images to extract feature point information from the at least one or more input tile images. The first interface unit divides the first source image by using the respective parts of the first source image on the basis of the horizontal resolution of the first source image to generate the at least one or more tile images.