Novel hybrid image content descriptor
D Savitha, Malay Kumar Nema · 2015
Efficient descriptors perform a key role in the success of object recognition, scene classification, image retrieval and many other computer vision related tasks which may learn the descriptors to decide on favorable class. In this paper we have proposed a hybrid descriptor which leverages the benefits of color as well as edges features. We extract color histogram from H component of the mean-shift filtered image. For edge features, we apply Primitive Set (PS) description on I component. The mean shift filtered color descriptor facilitates the identification of similar color images and the PS Descriptor adds value from its strict pass of similar structural elements. The usage of PS adds resiliency to the noise and blur to the descriptor. The results are conforming to the expectations and shown with the help of a retrieval system on Corel data-set.