Beauty and beast: image retrieval for image remodeling
Xuelong Li · 2004
This work presents a novel image retrieval for image remodeling (IR4IR) technique. A traditional content-based image retrieval (CBIR) system is regarded as a basis and colour features are selected to present the images' contents. After retrieving the related results, top positive examples are chosen to remodel the submitted query image by splitting an image into some minor sub-blocks and replacing them, meanwhile, the shifting information of each sub-block can be embedded into the least-significant-bit (LSB) of the corresponding pixels' colour channels, so that the remodeled image can come back to the original one. The new technique is helpful for arts related research and secures communication. In the experiments, the pre-process of image retrieval is on nearly 60,000 images from the Corel database, and the experimental results of IR4IR clearly show the advantage of the new methods for its application.