Image Retrieval Technology Research Based On The Spatial Relations Of Feature Points

This paper, image retrieval algorithm based on the spatial relations of featurepoints, which is based on the image low-level characteristics on research, is animportant branch of the content-based image retrieval algorithm. The low-level of theimage feature extraction and description is seen as the first step in content-based imageretrieval technology, which is also been regarded as one of the key technologies, andwhether the feature extraction and description method is good or not is closely related tothe performance of retrieving images quality. In addition, in view of content-basedimage retrieval application prospect and commercial value, it is significance andnecessary to do some research on this topic.In this paper, we put our focus on how to take the spatial relationship of featurepoints into full consideration for the purpose of good image description. Before this, thedevelopment of image retrieval technology is reviewed, and the domestic and foreignresearch present situation is introduced and so is the feature points based on spatialrelationship description method. Armed to solve the weaknesses of image retrievaltechnology which is based on spatial relationship of feature points, the concrete analysisand development trend research is well done, and then two kinds of image retrievalalgorithm are put forward.The first algorithm is that, in view of the deficiency annular histogram can’t givefull consideration to the space distribution of the feature points, we propose an imageretrieval method which is first to do image coarse matching by the annular histogrammethod, and then use Delaunay triangulation network, which can accurate describe thefeature points’ spatial relations, to do precise matching. Finally, apply it into trademarkimage.The second image retrieval algorithm, we propose, which is based on Hu invariantmoment and improved annular histogram to solve the problem that DT networkdescription is too accurate in the trademark image retrieval algorithm based on spatialrelationship of feature points. The standard deviation ellipses, which is the concept ofGIS is used to describe the space distribution of the images, so as to make the generalitymore batter.