Research On Image Retrieval Algorithm Based On Hybrid Feature

Image retrieval has always been a hot topic in computer graphics.As the use of mobile phones and the Internet increasingly frequent,people often come into contact with the image data also exploded.Image retrieval for specific images has also become an important application.The traditional image-based retrieval method based on text annotation is composed of a large number of manual work,subjective and too cumbersome.More and more people directly extract the image content characteristics,and calculate the similarity between images A research hot spot.Color is the most important visual perception of the world,the color characteristics of the stability is good,but it is often difficult to express the spatial characteristics of the image.Shape features describe the image by describing the contours or variations of the image,but the shape features may not be consistent with the intuitive perception of the human eye and are susceptible to noise.To this end,this paper presents a hybrid feature based image retrieval algorithm.This paper proposes a color feature extraction algorithm that includes position space information.By dividing the image into the space,the image is expressed as a partition color histogram feature of multiple regions in the HSV color space,which includes the global color feature and,to a certain extent,Location Features.At the same time,a hybrid shape feature extraction algorithm based on edge extraction is proposed.Based on the edge extraction,the Fourier shape descriptor,the eccentricity and the density shape descriptor are used to express the shape features of the image.Finally,we use the weighted color and shape feature similarity calculation method to calculate the similarity between the original image and the query image.In the operating system for the 64-bit Win10,the processor for the i7,8G memory for the machine,the use of Matlab R2014 a programming algorithm.The image training set and the test set are constructed by extracting some images from the Corell image data set.Experiments show that the algorithm is feasible and the expected effect can be achieved.