Image inquiry method based on clustering

Abstract

The invention discloses an image inquiry method based on clustering and includes the following steps: (1) extracting color characteristics and grain characteristics of target images and each image in an image database; (2) reducing the dimensions for high-dimensional characteristic to describe image contents, so as to gain characteristic subsets; (3) clustering data sets together to form characteristic subsets; (4) establishing indexes for each cluster in the clustering respectively; (5) using the index to perform image inquiry. The invention has the advantage that simple and effective partition can be performed for images according to concerned areas in images, so as to optimize the index tree structure and make the searching more accurate and effective.

Description

technical field [0001] The invention relates to image content representation and image query technology, in particular to a clustering image query method in high-dimensional data space. Background technique [0002] With the rapid development of the network, various types of information are increasing rapidly. In addition to text, the Internet (Internet) is also constantly producing a large amount of visual data such as images and videos. Image has become an important information carrier and an important content of multimedia because of its intuitive and vivid features and rich content. According to people's growing demand for visual data, how to effectively analyze, manage, query and retrieve these massive information has become an urgent problem to be solved. [0003] The traditional retrieval system mainly builds an index through the file name of the image and some text information of the webpage, and describes and retrieves the image. However, manual annotation is too...

AI Technical Summary

Problems solved by technology

[0007] Most tree-type spatial indexes have superior performance in low-dimensional space, but in high-dimensional space, the performance decreases, because some important parameters, such as volume, area, etc., have a power-level growth relationship with the spatial dimension

The R* tree has an ideal processing effect on data sets below 10 dimensions, but when the dimension exceeds 20, there will be a large number of MBR overlaps in high-dimensional spaces, resulting in structural degradation and performance degradation

Moreover, the data sets extracted in real-world applications, such as feature descriptors extracted from retrieved images and video key frames, are often non-uniformly distributed in the entire feature space, which will lead to a large number of empty nodes or sparse bounding rectangles. , seriously affecting the retrieval efficiency

Images