Abstract: Partial duplicate image revival plays an ample role in the real world application such as landmark search, copy protection, fake image detection. Users regularly upload images which are partially duplicate images on the domains like social site Facebook, hike and whatsapp etc. The partial image is only part of whole image, and the various kind of transformation like scale, resolution, illumination, and viewpoint. This technique is demanded by various real world applications and this has been motivated towards this research. In object based image retrieval methods we use the complete image as the query. This revival technique is similar to object based image retrieval. This technique is compare with image revival system by using the bag of visual words (BOV). Typically no any spatial information is used to retrieve image, so this approach is not execute in background noise. There is lots of background noise in the images and impossible to perform operation on the large scale database of the images.Two observations are. First, public show various objects through the images which are shared on the web, we also hope that the returned result also focus on the major parts or objects. Regions of interest are only found in salient region of the revival. The similar region in the returned result also identical to the salient region of the images. To filter out the nor-salient region from the image, which able to eliminate the background noise we introduce visual attention analysis technique. We also generate saliency region which having the expected visual contents.
Keywords: Partial duplicate image, Bags of visual words, Visual attention, Saliency feature, visually salient and rich region (VSRR) Query image, Source image, Sparse coding.