A technique to detect books from library bookshelf image

In this paper, we present a technique to recognize books from library bookshelf images. The proposed technique includes several modules such as - noise removal, bookshelf row detection and book detection and extraction from bookshelf. Firstly, the image is filtered for noise removal and edges are detected. The proposed technique counts the number of pixels of every candidate for horizontal lines on the edge detected image and compares the value with a threshold. Elected lines are extended which creates some horizontal lines segmenting the image into different portions. These individual portions are considered as bookshelf row regions of the bookshelf and each of them is passed into the book detection module for further processing. Applying vertical line detection in the manner of horizontal line detection method on the segmented regions is not adequate to detect a book from shelf row because of the angular orientation of book arrangements. The proposed technique detects books arranged in different angular orientation on a shelf and differentiates between the book and non-book components based on some properties of the book shape. Experimental results show that the proposed technique is capable of extracting bookshelf rows and detecting individual books from library bookshelf image successfully.