A non-smooth non-local variational approach to saliency detection in real time

In this paper, we propose and solve numerically a general non-smooth, non-local variational model to tackle the saliency detection problem in natural images. In order to overcome the typical drawback of the non-local methods in image processing, which mainly is the inherent computational complexity of non-local calculus, as the non-local derivatives are computed w.r.t every point of the domain, we propose a different scenario. We present a novel convex energy minimization problem in the feature space, which is efficiently solved by means of a non-local primal-dual method. Several implementations and discussions are presented taking care of the computing platforms, CPU and GPU, achieving up to 33 fps and 62 fps respectively for 300x400 image resolution, making the method eligible for real time applications.

Automated summary

In this paper, a general non-smooth, non-local variational model is proposed and numerically solved for saliency detection in natural images. By formulating a convex energy minimization problem in feature space and employing a non-local primal-dual method, the computational complexity is significantly reduced while achieving high performance for real-time applications on both CPU and GPU platforms.