An Image Retrieval Method Based on Clustering Distance Direction Histogram

Abstract

The invention discloses an image retrieval method based on a clustering distance direction histogram, belongs to the technical field of pattern recognition and information processing, and relates to massive image retrieval in computer vision. This method counts the distribution characteristics of the local visual features extracted from the training image set, and divides the local feature space; through the descriptor of the local feature of the target image and the distance between the center of the divided space, and the distance between the main direction of the local feature and the main direction of the divided space The consistency of local features is assigned to the nearest sub-region, and then the full-effect expression of the image is generated; using the full-effect expression of the image for retrieval can efficiently complete the retrieval of large-scale images. Through a large number of experiments, it is verified that the present invention effectively improves the accuracy of image retrieval and obtains good scalability on large-scale image sets under the condition of using less calculation.

Description

technical field [0001] The invention belongs to the technical field of pattern recognition and information processing, and relates to massive image retrieval in computer vision. Background technique [0002] With the popularization and development of the Internet and digital photography equipment, images on the Internet have exploded. How to quickly and efficiently obtain images of interest from massive image data is particularly important, attracting the attention of more and more experts and scholars. As the scale of images increases, the difficulty of image retrieval also increases accordingly, which not only requires high retrieval accuracy and efficiency, but also requires as low storage overhead as possible, and at the same time, it is necessary to ensure the scalability of the image retrieval system. Therefore, the large-scale image data and the diversity of its contents increase the need for compact image representation with high recognizability. In order to obtain...

AI Technical Summary

Problems solved by technology

However, during the quantization process of VLAD descriptors, the quantization errors of local descriptors are positive and negative, and may cancel each other out.

At the same time, the VLAD descriptor only uses the descriptor information of local features, while ignoring other useful clues, such as spatial information, which affects its retrieval performance and scalability

Images