A genetic-based prototyping for automatic image annotation. (August 2018)

• Record Type: Journal Article
• Authors: Maihami, Vafa; Yaghmaee, Farzin
• Is Part Of: Computers & electrical engineering. Volume 70(2018)
• Page Start: 400
• Page End: 412
• Publication Date: 2018-08
• Subjects: Automatic image annotation -- Genetic algorithm -- Prototyping -- Relevance tags -- Image retrieval; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2017.03.019
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlight: We proposed a novel method for automatic image annotation. An effective framework is introduced to solve the number of images in the nearest-neighbor-based image annotation. A genetic-based algorithm is proposed to find the best prototypes of training images. Abstract: The explosive growth of visual and textual data has led to urgent requirements in management and understanding of digital content. Developing optimal solutions to allow access to and mining such data in modern applications is crucial. Image annotation or tagging, is a process which produces words, keywords or comments to an image. In the nearest-neighbor-based automatic image annotation, a training set T is given to a classifier for classifying new prototypes. In practice, T contains useless images for the image annotation task, that is, superfluous prototypes, which can be noisy or redundant; therefore a process is needed to discard them from T . In this paper, a genetic-based prototyping for automatic image annotation is proposed. We first adopt a genetic-based prototyping algorithm to obtain optimal prototype from images. Then, for a given query image, its neighbor images are retrieved from the optimal prototype gained, and to generate its candidate tags some methods such as voting are used. Experimental results on standard benchmark datasets show that the proposed method achieves order of magnitude speedups over the related techniques and obtains much better annotate quality as well. Graphical abstract: