Multi label learning and multi feature extraction for automatic image annotation

摘要

Recently, various multimedia technologies has been developed, which increase the collection of digital images. In daily life, popularity of digital camera and social media is also increased which is resulted in huge digital data sharing. Within such large amount of image data, specific image searching is very difficult. To make ease of searching, dictionary learning becomes popular solution. The feature based image annotation is the new area for image searching. In this image annotation task, some human keywords are assigned to the images, so that searching becomes easy. In this paper, we present a multi-label learning and multi keyword extraction for automatic image annotation. This framework is worked in two phases named as training and testing phase. In training phase, we build the classifier with the help of extracted features, mapping of tags and features and dictionary learning. This classifier is used to identify the labels for testing image. For classification we have used C4.5 classifier and prove that the accuracy and efficiency is better than naïve byes classifier. The performance of system is tested on IAPR TC12 dataset. Experimental results prove that the multiple label and multiple features extraction improves the efficiency of image annotation framework.