Semantic visual dictionary optimization method based on Pearson correlation coefficients

Abstract

The invention discloses a semantic visual dictionary optimization method based on Pearson correlation coefficients. The method comprises the following steps: firstly, extracting bottom layer features of colors, shapes and textures of an image, and extracting dynamic features of the image by using an SURF algorithm; clustering the acquired image dynamic features by adopting an E2LSH clustering algorithm, extracting related description visual phrases, and constructing an original visual dictionary; and introducing a Pearson correlation coefficient to solve the correlation degree between the dynamic features and the correlation degree between the bottom layer features and the dynamic features so as to optimize the original visual dictionary and obtain a final semantic visual dictionary. In order to solve the problems that visual phrases in a semantic visual dictionary are redundant and calculation is complex. Images are classified so that the classification performance of the images is improved. Operation complexity is reduced. Operation time is shortened, and classification accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of visual dictionary optimization, in particular to a semantic visual dictionary optimization method based on Pearson correlation coefficient. Background technique [0002] For a long time, image classification technology has attracted the attention of many scholars at home and abroad, and its market and social application value have also been affirmed by people. Different from text information, image and video information have greater information volume and more difficult to understand content. It is still more difficult to make a computer understand an image and a video like a human being. In real life, people obtain the information contained in the image through the eyes, and then process the obtained information through the brain. The human brain will remove the noise or useless information and retain the corresponding image information. When encountering an image of the same type next time, the brain w...

AI Technical Summary

Problems solved by technology

However, an excessively large visual dictionary will increase the time complexity of image classification, and to classify or recognize an image, it is inevitable that some image noise will affect the classification or recognition efficiency.

These image noises will not only reduce the classification accuracy of the image, but also increase the size of the visual dictionary. An excessively large visual dictionary will increase the time cost of computer classification.

Images