Characterizing noisy or ancient documents is a challenging problem up to now.Many techniques have been done in order to effectuate feature extraction andimage indexation for such documents. Global approaches are in general lessrobust and exact than local approaches. That's why, we propose in this paper, ahybrid system based on global approach(fractal dimension), and a local onebased on SIFT descriptor. The Scale Invariant Feature Transform seems to dowell with our application since it's rotation invariant and relatively robustto changing illumination.In the first step the calculation of fractal dimensionis applied to images in order to eliminate images which have distant featuresthan image request characteristics. Next, the SIFT is applied to show whichimages match well the request. However the average matching time using thehybrid approach is better than "fractal dimension" and "SIFT descriptor" ifthey are used alone.
展开▼
