Low quality character images include a lot of variousness and noises, so the use of a single classifier is difficult to achieve a high recognition performance. For this, we experimented various feature extraction methods suitable for low quality image...
Low quality character images include a lot of variousness and noises, so the use of a single classifier is difficult to achieve a high recognition performance. For this, we experimented various feature extraction methods suitable for low quality images and constructed a classifier-pair that uses two complementary features. The experimental result using this classifier-pair was very effective for the recognition of low quality character images.