The sensor values captured by a digital camera are transformed in a non-linear manner prior to quantization in order to make the quantization rate approximately proportional to the sensitivity of the human visual system. We propose an image dependent non-linear transformation that can accurately reproduce the detail and contrast visible in the original scene. The principles underpinning the transform stem from an understanding of natural image statistics, as well as recent experimental and neurophysiological findings. To optimize the parameters of the model we collect user-feedback and develop a method that can predict the user defined parameters. The method we have developed has an extremely low computational complexity, therefore it operates almost instantaneously making it suitable for in-camera operations. The final image looks natural, without any halos, spurious colors or artifacts. It can also be applied to video sequences, after imposing temporal coherence on the parameter values by smoothing them over time. The proposed approach is validated through psychophysical tests that confirm that it outperforms other state of the art algorithms in terms of users' preference.