A novel image processing approach combining a 'coupled nonlinear oscillators'-based paradigm with cellular neural networks for dynamic robust contrast enhancement

In this paper, a systematic discussion of both pros and cons of two well-known traditional approaches for image contrast enhancement is conducted. The first approach is based on the CNN paradigm and the second one is based on the coupled nonlinear oscillators' paradigm for image processing. In the later case an extensive bifurcation analysis is carried out and analytical formulas are derived to define the various states of the system. Both equilibrium and oscillatory states of the system are depicted. It is shown that each of these states has a significant impact on the quality of the resulting image contrast enhancement. A benchmarking is considered whereby a comparison is performed between the results obtained by a CNN-based processing, on one side, with those obtained by a 'coupled non linear oscillators' based processing, on the other side. The superiority of the later approach (for contrast enhancement) is demonstrated both analytically and through various experiments. A major drawback of the CNN based image processing is the practical inability to adjust/re-calculate templates in real-time in face of a dynamic scene with input images experiencing visibility and/or lighting related spatio-temporal dynamics. Finally, a novel hybrid approach integrating both schemes in an efficient way is proposed: the 'coupled nonlinear oscillators' based image processing is the main processing scheme that is however realized on top of a CNN processors' framework. The hybrid approach does prove to overcome key practical problems faced by both original approaches.