Content based image indexing and retrieval using a textual approach in visual online shopping.

In this study a new Content-Based Image and Indexing Retrieval engine based on the textual representation of the extracted features is proposed. We discuss about the advantages in the use of the proposed method both from the point of view of the quality of the results than the efficiency and the ease in the integration of the visual features in a real context. Furthermore, one of the main aspects of the proposed method consists in the use of an object detection phase which is essential to extract the visual information only from the Object of Interest within the image. The main contributions of this work consist in the development of a new color feature based on the Human Visual Perception able to extract a limited set of representative colors from the whole RGB space, and the development of a new feature able to extract the Visual Attributes in the domain of fashion based on the Human Pose estimation. Moreover a prototype based on the results of this study has been proposed, which is designed as an engine for the online shopping, focusing on the fashion domain which is the most promising field to test the extraction of visual information. As an additional contribution of this study, given the increasing amount of mobile devices and the consequently interest in the mobile shopping, an extension of the proposed prototype has been developed as a mobile application able to perform a simple query by similarity.