The Study Of Image Deduplication And The Choice Centroid-Image

This paper mainly researches the unresolved issues about image de-duplication. image de-duplication has two classification, for the second class of problems we put forward the corresponding solution, by giving the definition of graffiti images and image noise with distinction and connection. For the second image de-duplication questions, combine and compare the sequential DCT-based image copy algorithms and FAIDA to find the improved filtering algorithm and Hash centroid choose to avoid algorithms. Collection of images for image de-duplication question, the image matching technology to de-duplication areas of the image, first proposed an image based on image matching algorithms de-duplication,the proposed solution image cropping problem sets.Improved Hash filtering algorithm in FAIDA Hash filtering algorithm, based on the concept of image graffiti index, will meet graffiti image index threshold required image into the same bucket for subsequent processing. In order to avoid the type of graffiti image selection image of the center of mass, we proposed to avoid the centroid algorithm selection, the choice of pre-processing before the image centroid to exclude the interference of graffiti images.Compared with a collection of graffiti images, image cropping sets can not be a quick way to find duplicate images filtration. This paper proposes an image based on image matching eliminate redundant algorithm, which is the image matching algorithm as the core of the image eliminate redundant steps, by raising the efficiency of image matching algorithm to achieve the image FMCG redundant.This paper gives a detailed step above algorithm, flowchart and pseudo code, and the feasibility and advantages and disadvantages of the algorithm.The work on the future outlook, will solve problems more images eliminate redundant.