Calligraphy image style classification method based on direction feature enhancement

Abstract

The invention discloses a calligraphy image style classification method based on direction feature enhancement, and the method comprises the steps: extracting original images of different calligraphy styles, dividing the collected original images into a training set and a test set, carrying out the direction feature extraction of eight directions of each image in the images of the training set through a 2D-Gabor filter, and carrying out the classification of the styles of the calligraphy images. And then enhancing the original image by using the extracted direction features, inputting the enhanced original image into the deep convolutional neural network for training, testing the deep convolutional neural network on a test set every time training for a certain number of times to obtain test accuracy, and storing a weight parameter corresponding to the highest test accuracy to obtain a weight parameter corresponding to the highest test accuracy. And forming a final calligraphy style classifier with the deep convolutional neural network. According to the method, the 2D-Gabor filter is adopted to extract the direction feature mapping of the calligraphy image as priori knowledge to enhance the feature representation of the original image, so that the method not only has higher classification accuracy, but also has better generalization ability.

Description

technical field [0001] The invention belongs to the field of artificial intelligence, and in particular relates to a calligraphy image style classification method based on direction feature enhancement. Background technique [0002] Chinese calligraphy works are cultural treasures and have important research value. In the history of the development of Chinese culture, countless calligraphic fonts with rich connotations have been left behind. Four famous masters of regular script in ancient my country: Ouyang Xun, Yan Zhenqing, Liu Gongquan, Zhao Mengfu. The calligraphy styles created by them: European style, Yan style, Liu style and Zhao style have far-reaching influence, and are still loved and copied by many people. [0003] These four calligraphy styles belong to regular calligraphy in essence. There is little difference in the characteristics of strokes between them, which is difficult for untrained people to distinguish. It is quite labor-intensive to find experienc...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the above-mentioned deficiencies, provide a kind of calligraphy image style classification method based on directional feature enhancement, solve the problem of low accuracy rate of calligraphy style classification existing in the prior art, and improve the generalization ability of the classification model

Images