Content based image retrieval using local and global features descriptor
Ali Douik, Mehrez Abdellaoui, Leila Kabbai · 2016
Recently, Content Based Image Retrieval (CBIR) has received a great attention by researchers. It becomes one of the most interesting topic in computer vision and image processing. CBIR image can be represent by local or global features. The entire image is described in the case of global features by using a novel descriptor called Upper-Lower of Local Binary Pattern (UL-LBP) based on Local Binary Pattern (LBP). Whereas, local features extract the Interest Points (IP) using Scale Invariant Feature Transform algorithm (SIFT). These features take into account the color channels information (Red, Green and Blue) independently in order to enhance results. This paper presents a hybrid approach for CBIR which combines both local and global feature of an image to generate a new descriptor denoted Histogram of Local and Global features using SIFT (HLG-SIFT). The performance of our descriptor is evaluated by computing the precision and recall using Euclidean distance and compared to state of the art.