A Deep Study of Content Based Image Retrieval System using Sentiment Analysis

Analysis of visual contents has always been interesting and important yet it is very challenging as well. With the increasing popularity of social grids, images are considered a very expedient way to communicate and diffusion of information among online users. To know the different patterns and different aspects of these images it is very important to first interpret these images in a simpler form. Like the textual information images also carry different levels and different types of sentiments to their spectators. Though it is quite easy to detect any type of sentiment from the text but it is very difficult to analyse sentiments from the visual images. By using the CBIR technique it would be quite easy to get the accurate image but the image with right sentiment is again a challenge. In this paper, I have presented a method which is based on psychosomatic models and web mining that can easily and automatically construct a huge set of Visual Sentiment Ontology (VSO) which comprises around 4000 ANP (Adjective Noun Pairs). I have also proposed the concept of SentiBank, a pictorial notion sensor library that can be used to sense more than 1000 ANPs in an image. These two technique, VSO and SentiBank will positively open the new doors to analyse the sentiments in an image and gives the more accurate results while accessing the images using sentiments.