Images of high (or) super resolution will be good to see. And having excellent clarity in all kinds of devices. Nowadays all mobile displays are not providing the users expecting clarity. Snaps taken from the mobile will be blurred (or) not clear sometimes. Reason for these problems are not using clarity lenses, shaking of mobile while taking snaps. For getting clarity in images first we should consider the pixels of the images. Pixels of an image should be very minute. If the image having very less spilt up in their pixels, then that image will look like blurred. So separation of pixels in images plays an important role here. For considering the pixels we should first use manifold algorithm. It split the images as pixels with good clarity and split as according to their texture also. In contrast to existing methods which train a binary classifier for each keyword, our keyword model is constructed in a straightforward manner by exploring the relationship among all images in the feature space in the learning stage. In relevance feedback, the feedback information can be naturally incorporated to refine the retrieval result by additional propagation processes. In order to speed up the convergence of the query concept, we adopt two active learning schemes to select images during relevance feedback. Furthermore, by means of keyword model update, the system can be self-improved constantly. The updating procedure can be performed on-line during relevance feedback without extra off-line training.
Prof. Dr. Bilal BİLGİN