Social computing for image matching

• Published in: PloS one Vol. 13; no. 5; p. e0197576
• Main Authors: Chamoso, Pablo, Rivas, Alberto, Sánchez-Torres, Ramiro, Rodríguez, Sara
• Format: Journal Article
• Language: English
• Published: United States Public Library of Science 29.05.2018; Public Library of Science (PLoS)
• Subjects: Added value; Algorithms; Archives & records; Artificial Intelligence; Augmented reality; Big Data; Causality; Computer and Information Sciences; Data analysis; Data processing; Forecasting; Humans; Image detection; Image processing; Image Processing, Computer-Assisted - methods; International conferences; Internet; Matching; Medicine and Health Sciences; Methods; Online Social Networking; Physical Sciences; Probability; Problems; Research and Analysis Methods; Social networks; Social organization; Social Sciences; Trends; User-Computer Interface; Spain
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0197576

One of the main technological trends in the last five years is mass data analysis. This trend is due in part to the emergence of concepts such as social networks, which generate a large volume of data that can provide added value through their analysis. This article is focused on a business and employment-oriented social network. More specifically, it focuses on the analysis of information provided by different users in image form. The images are analyzed to detect whether other existing users have posted or talked about the same image, even if the image has undergone some type of modification such as watermarks or color filters. This makes it possible to establish new connections among unknown users by detecting what they are posting or whether they are talking about the same images. The proposed solution consists of an image matching algorithm, which is based on the rapid calculation and comparison of hashes. However, there is a computationally expensive aspect in charge of revoking possible image transformations. As a result, the image matching process is supported by a distributed forecasting system that enables or disables nodes to serve all the possible requests. The proposed system has shown promising results for matching modified images, especially when compared with other existing systems.