GAN‐based tone curve learning for colour transfer

A new approach for reflecting the colour tone of a reference image on input is proposed. Depending source and pairs, conventional statistical transfer methods often lead to undesirable transfer. Conversely, deep learning depend prior learning, which results in unnatural output images when inappropriate are learned; moreover, such situations, analysing what kind transformation has actually been performed difficult. This state art motivates proposal method that estimates curves based generative adversarial nets. does not require any data set other than images, thus enabling more appropriate The superior proposed compared with some baseline approaches demonstrated through experiments.