In recent years, a large number of binarization methods have been developed,with varying performance generalization and strength against differentbenchmarks. In this work, to leverage on these methods, an ensemble of experts(EoE) framework is introduced, to efficiently combine the outputs of variousmethods. The proposed framework offers a new selection process of thebinarization methods, which are actually the experts in the ensemble, byintroducing three concepts: confidentness, endorsement and schools of experts.The framework, which is highly objective, is built based on two generalprinciples: (i) consolidation of saturated opinions and (ii) identification ofschools of experts. After building the endorsement graph of the ensemble for aninput document image based on the confidentness of the experts, the saturatedopinions are consolidated, and then the schools of experts are identified bythresholding the consolidated endorsement graph. A variation of the framework,in which no selection is made, is also introduced that combines the outputs ofall experts using endorsement-dependent weights. The EoE framework is evaluatedon the set of participating methods in the H-DIBCO'12 contest and also on anensemble generated from various instances of grid-based Sauvola method withpromising performance.
展开▼
