In this paper, we propose a novel method for automatically
predicting the impact of the quality of the cameracaptured
image on OCR performance without requiring a
reference image (i.e. a no-reference method). Unlike other
image quality assessment methods, which only identify some
certain blurs and their parameters, or predict the SNR-related
scores that do not have a simple correlation between the
image quality and its applicability to the document analysis
system, the proposed method measures the quality or degradation
level of the document image according to its predicted
OCR score and accepts/rejects images accordingly.
We organize the rest of the paper as follows. Section 2 introduces
the feature extraction method on document images.
Section 3 describes the Support vector machine (SVM) based
regression and classification method which predicts the document
degraded score. Section 4 presents the details of the
experimental set up and results. The conclusions are covered
in Section 5.