Text image retrieval based on generalized normalized picture information measure

The gap between statistical feature and semantic feature of data information is rather large. And it´s difficult for computers to retrieve text image based on semantic feature directly. So the performance of text image retrieval algorithm depends on statistical feature abstraction to great extent. To this question, the generalized normalized picture information measure (GNPIM_k) is introduced as the statistical feature. Experiment results show that the text image retrieval algorithm based on GNPIM_k is effective.