Training of Neural Networks for Image Correction with Natural Images

G. Krell and B. Michaelis (Germany)


Image correction, neural networks, image restoration, image enhancement


The front-ends and back-ends of digital image processing systems are acquisition and reproduction devices respectively. A system of correction modules is presented in this paper that can correct for errors in both image acquisition and image reproduction devices. This improves subjective and objective image quality, further benefiting subsequent pattern recognition and image analysis stages. The deployed modules are artificial neural networks trained with natural images that compensate for errors in the image acquisition and reproduction devices.

Important Links:

Go Back