Ancient Chinese Character Image Segmentation Based On IVHFS And IDFA

The ancient Chinese character image segmentation is the process of segmenting the ancient book layout images to obtain a single ancient Chinese character image.It is a key link in the ancient Chinese characters image recognition and retrieval.The traditional method of character segmentation has a low accuracy rate in the ancient Chinese character image segmentation,and thereby affects the subsequent ancient Chinese character image recognition and retrieval.This is mainly due to the complex structure of ancient book layout and the diverse styles of ancient Chinese characters.Therefore,it is necessary to study an effective ancient Chinese character image segmentation algorithm based on the characteristics of the ancient book layout and ancient Chinese characters to meet the need of subsequent ancient Chinese character image recognition and retrieval.This study has taken "Si Ku Quan Shu" of Wenyuan Pavilion layout image as the research object,and researched on the ancient Chinese character image segmentation algorithm.Based on the preliminary segmentation of the ancient book layout image,follow-up processing for various types of segmentation errors are carried out separately.Related work includes:(1)An over-segmentation algorithm of ancient Chinese character images based on IVHFSOver-segmentation ancient Chinese character images refer to the situation where the segmentation results only contain a certain Chinese character partial stroke due to the particular structure of Chinese characters and other reasons,and the images need to be merged to obtain a complete ancient Chinese character image.Taking the unique advantage of interval-valued hesitant fuzzy set(IVHFS)in handling uncertain problems,this study considered the combination of over-segmentation Chinese character components synthetically,performed multi-attribute evaluation,and defined evaluation functions.The interval value hesitation fuzzy set is established,and the distance measurement of other Chinese character components to the over-segmentation Chinese character component is calculated,and the over-segmentation Chinese character components are merged iteratively,thereby obtaining the complete ancient Chinese character image.(2)An under-segmentation algorithm for ancient Chinese character images based on IDFAUnder-segmentation ancient Chinese character images refer to the situation where there are more stroke pixels than a single Chinese character in the segmentation result due to various reasons such as adhesion,and further segmentation processing is required to obtain a single ancient Chinese character image.Use an improved drop fall algorithm(IDFA)to deal with Under-segmentation adhesive ancient Chinese character images.The traditional drop fall algorithm has shown high accuracy in dealing with the problem of adhesive characters.This study improved the dropping rule of drop fall algorithm to be suitable for the situation of adhesive ancient Chinese character images,and used K-means algorithm to cluster pixels of under-segmentation ancient Chinese character images with high cohesion,so as to determine the initial drop point,thereby solving the problem of under-segmentation ancient Chinese character images.This study used the ancient book database in "Si Ku Quan Shu" of Wenyuan Pavilion commonly used in the study of ancient Chinese characters as the experimental data set,and randomly selected 100 layout images which contain a total of 28110 ancient Chinese characters.The accuracy rate of the over-segmentation Chinese character processing algorithm was 87.12%,the experimental results showed that the accuracy rate of the algorithm was 89.42%,and the accuracy rate of the ancient Chinese character segmentation was 89.94%.The proposed method in this study can adapt to the characteristics of ancient books and improve the accuracy of ancient Chinese character image segmentation.