Binary image feature extraction method and system

Abstract

The present invention is applicable to the technical field of image retrieval, and provides a binary image feature extraction method. The method includes: obtaining the position of feature points according to preset rules; selecting several pixels as binary comparison points; taking each binary comparison point as center, perform Gaussian blur processing on the binary comparison points; from all the binary comparison pairs formed by the binary comparison points, select several pairs of comparison pairs with the largest discrimination and the least correlation; select several pairs of binary comparison pairs , performing pixel attribute comparison to generate a binary descriptor; converting the binary descriptor into an integer according to a preset rule. The present invention also provides a binary image feature extraction system for realizing the above method. Thereby, the present invention refers to the imaging principle of the human retina, so that the efficiency of the subsequent image matching retrieval algorithm is greatly improved.

Description

technical field [0001] The invention relates to the technical field of image retrieval, in particular to a binary image feature extraction method and system. Background technique [0002] The rapid growth of visual information such as Internet images has brought great challenges to the organization and management of information. There is an increasing demand for content analysis and detection of massive images. Image copy detection, image indexing, and content-based image search engines, etc. Technology came into being. Currently, in these technologies, local feature point (Local feature point) features are usually used as features of image content. As the name suggests, local features are features that only appear in some parts of the picture. When the object is not completely occluded, these local features can exist stably and have good distinguishability, so these local features can be used as the whole image. Characteristics. [0003] The extraction of local feature p...

AI Technical Summary

Problems solved by technology

Representatives of real-number descriptors mainly include SIFT, SURF, etc. Each feature component of a real-number descriptor is a real number, so the description range is wide and the degree of discrimination is high, but its calculation is large, and the calculation in the subsequent feature matching is complicated. The algorithm is difficult and not suitable for real-time computing

Representatives of binary descriptors mainly include ORB, BRIEF, BRISK, FREAK, etc. Since each feature component of binary descriptors is a simple 01 binary, although the discrimination of a single description component is not strong, due to its simple calculation and occupied storage The space is small, and multiple sets of feature components can be calculated to improve the descriptive power. However, the number of feature components will affect the efficiency of the subsequent feature matching algorithm. Therefore, how to balance the feature discrimination and algorithm efficiency is the difficulty of the local feature point descriptor extraction algorithm. , in addition, as the number of images in the image library increases, the matching speed of both real-number descriptors and binary descriptors will decrease significantly. How to make the decrease as small as possible is still the difficulty of the local feature point descriptor extraction algorithm

[0006] In summary, there are obviously inconveniences and defects in the actual use of the existing technology, so it is necessary to improve

Images