Image retrieval method and system based on k-nearest neighbor and fuzzy pattern recognition

Abstract

The invention discloses an image retrieval method and system based on k-nearest neighbor and fuzzy pattern recognition. The method comprises the steps that colors and texture feature vectors are extracted respectively aiming at query images and retrieved images, fuzzy normalization processing is conducted, and fusion is conducted on fuzzy colors and texture features to obtain comprehensive featurevectors of corresponding images; K near images of the query images are searched aiming at the obtained query images and the comprehensive feature vectors of all of the retrieved images; the similarity between the query images and the k near images is calculated, and the similarity among each retrieved image and the k near images of the query images is calculated to obtain corresponding k-dimensional fuzzy feature vectors of the query images and each retrieved image; the fuzzy similarity among the corresponding k-dimensional feature vectors of each retrieved image and the k-dimensional fuzzy feature vectors of the query images is calculated; the retrieved images are fed back to a user in the order from high to low according to the fuzzy similarity; whether or not the image retrieval process is stopped is judged according to the satisfying degree of the user.

Description

technical field [0001] The invention relates to an image retrieval method and system based on k-nearest neighbor and fuzzy pattern recognition. Background technique [0002] With the rapid development of multimedia technology and network technology, image data is increasing at an alarming rate every day. How to effectively and quickly retrieve images of interest to users from massive image databases is a very important and challenging research topic. . Currently widely used image retrieval methods are: text-based image retrieval (Text-based Image Retrieval, TBIR) method and content-based image retrieval (Content-based Image Retrieval, CBIR) method. [0003] The TBIR method is to use the method of text annotation to describe the content of the image, and form keywords to describe the content of the image, such as scenes and objects in the image. Image annotation can be done manually or semi-automatically using image recognition technology. When performing image retrieval, ...

AI Technical Summary

Problems solved by technology

The TBIR method is easy to implement, and there is manual intervention in the labeling, so its accuracy rate is relatively high, but the TBIR method also has serious defects due to manual intervention: first, in a large-scale image database, all It will consume a lot of manpower and financial resources to label the images, so the TBIR method is only suitable for small-scale image databases; secondly, it is difficult to summarize the content of an image with precise and short words, which makes it difficult for the TBIR method to achieve accurate Inquire

In order to improve the effectiveness of the CBIR method, many methods for image feature extraction and image feature matching have been proposed, but these methods generally have the characteristics of cumbersome calculations, long retrieval time, and the retrieval effect is still unsatisfactory.

Images