Content Based Image Retrieval Based on Colour and Texture Features Using Hog Descriptor
Prashant Nair, Dhileep Kumar, Ramesh Shahabadkar · GLOBAL JOURNAL FOR RESEARCH ANALYSIS · 2012
Content based image retrieval had been a prominent research field. As far as the need of people using digital images is been increasing enormously, there has been an increased need for study and extension for image databases. Lot of interest had been given in retrieving the images from the databases. So, an efficient way to do the same has become a specific requirement. Thus an efficient algorithm should be made out to do the same. The images have to be characterized with certain features in order to identify an image. The basic visual features are the colour and the texture features. Therefore, an algorithm which uses the colour and the texture features has been proposed. Initially the image in the database and query image are partitioned into 6 equally sized tiles. Colour feature is represented by HSV histogram. The texture features is obtained by grey level co-occurrence matrix(GLCM). The colour feature is added to histogram of gradient(HOG) features used for object detection. A one to one matching algorithm is used to find the similar images. A threshold number of similar images are retrieved. In our paper we have set the threshold as 9 images out of the given dataset. Euclidean distance is used to compute the similarity distance. The experiments are reflected with their results to show the efficiency. ABSTRACT