Enhanced illumination balancing with Neural Network for improved degraded scanned text-photo images

Degradations in text-photo image result from poor quality of degraded scanned-documents, where considerable background noise or uneven illumination occurs. These disturbances may not be easily classified as objects or background pixels. This paper proposes the alternative illumination balancing based on the Neural Network that aims to improve the degraded scanned documents. The experiment results show that the proposed algorithm has produced a significant improvement in the performance. Moreover, the improvement can be clearly seen for text-photo images as the results were compared with other algorithms.