An inverse halftoning algorithm based on neural networks and UP(x) atomic function

Halftoning and inverse halftoning algorithms are very important image processing tools that have been widely used in digital printers, scanners, steganography and image authentication systems. Because such applications require obtaining high quality gray scale images from its halftoning versions, several inverse halftoning algorithms have been proposed during the last several years, which provide gray scale images with Peak Signal to Noise Ratio (PSNR) of about 25 to 28 dB. Although this may be enough for several applications, exist several other that require higher image quality. To this end, this paper proposes an inverse halftoning algorithm based on Upx atomic function and multilayer perceptron neural network. Experimental results show that proposed scheme provides gray scale images with PSNRs higher than 30dB independently of the method used to generate the halftone image.