Abstract: Image retrieval is a key challenges in many image database application and still an active field in computer vision application. There are many proposed image retrieval systems that retrieve images based on image contents such as colors, texture, shapes and feature descriptor. The main task for image retrieval system is to create a system that capable to retrieve images that are semantically related to user s query from an image database. When user interest to retrieve images that contain a particular objects instead of retrieve similar images which might not related images to his interesting this is called object based on image retrieval. The goal of Objects-based is to retrieve images based on objects that appear in those query images from large database. In this study we use enhanced Speeded UP Robust Features (SURF) algorithm as main step to extract features from interested query objects and then checked and matched result to retrieve related images from image dataset. Speeded UP Robust Features (SURF) is a scale and rotation invariant detector and descriptor feature algorithm and was applied successfully in many Image retrieval systems due its robust against different image transformation. Finally the result written in a report and images saved along with user s query time stamp.
Key words: Object Based Image Retrieval (OBIR), feature matching, object recognition, SURF, image database
INTRODUCTION market logs, cars, stop signs, flags and others where user
In recent years, the fast growth of image collections satisfy conditions such as: Object can distinguish from and databases has established a need for user-friendly background and has distinctive features like color o r image retrieval system that can retrieve images based on texture (Hoiem et al., 2005). Vision features can b e user s query. Image Retrieval system can be defined as the classified into two classes: first, low level features include system that find all images in a given image database that color, texture and inflexion. Second, middle level features depict scenes of some predefined user specification called that include shape description and object featur. queries (Bay et al., 2008). Image retrieval algorithms can In this study we used enhanced Speeded UP Robust be divided into two categories: text based and content Features (SURF) (Asadi and Obaid, 2016) one of the based approaches. Content Based Image Retrieval (CBIR) robust local feature detector and descriptor algorithm, first is an important field in image retrieval and is a complement presented by Bay et al. (2008). Speeded UP Robust approach to text based content. Content Based Image Features (SURF) is local feature descriptor of the Retrieval (CBIR) is an image retrieval system that i s interested points that detected in an integral image. These retrieve image based on its content by use low level visual descriptors can be used to detect the matching size content of an image like texture, shape, color and spatial between two images or for detection particular object information to represent and index the image and then while other objects exist in an image. For particular object measure similarity among images to return relevant result where user interest to find in an image, feature extracted based on the difference between low level features. that correspond to it should be similar even though are Many research work has been done in this field and many extracted in different illumination, scale and noise in order techniques were adopted to enhance the accuracy o f to perform accuracy in object detection and recognition. retrieval. Object Based Image Retrieval (OBIR) is a part of Speeded UP Robust Features (SURF) is well known image retrieval system and can be defined as the system algorithm that can be used in computer vision application that retrieve images from image database based on the that is invariant to scale, rotation and illumination change appearance of objects in those images. Objects can be and used to perform tasks of image retrieval, object