Similarity calculation method for images based on the scene graph

• Published in: Signal, image and video processing Vol. 17; no. 5; pp. 2395 - 2403
• Main Authors: Peng, Jinghui, Wang, Zhen, Wang, Shizhe
• Format: Journal Article
• Language: English
• Published: London Springer London 01.07.2023; Springer Nature B.V
• Subjects: Computer Imaging; Computer Science; Correlation coefficients; Image Processing and Computer Vision; Mathematical analysis; Multimedia Information Systems; Original Paper; Pattern Recognition and Graphics; Signal,Image and Speech Processing; Similarity; Vision; Scene graph; Dataset EVR; Triple-pairs similarity; Image similarity; Entities relationships
• ISSN: 1863-1703; 1863-1711
• DOI: 10.1007/s11760-022-02456-0

Many image similarity calculation methods have been applied to effectively manage and use image data. The problem is that the current image similarity calculation method does not eliminate redundant information and lacks the relationship between entities. This paper proposed a scene graph calculation method for image similarity. Firstly, analyzed and summarized the advantages and disadvantages of common image similarity calculation methods; secondly, combined the characteristics of the equipment scene graph to eliminate redundant information, clarified the interaction relationship between entities and correlation coefficient representation, and given a method for similarity calculation of equipment scene graph. Finally, designed and conducted experiments on similarity calculation on equipment visual relationship dataset with typical scene graph characteristics. Experiments show that the proposed method performs well in improving image similarity calculation.