Edge-Preserving Piecewise Linear Image Smoothing Using Piecewise Constant Filters

Abstract: Most image smoothing filters in the literature assume a piecewise constant model of smoothed output images. However, the piecewise constant model assumption can cause artifacts such as gradient reversals in applications such as image detail enhancement, HDR tone mapping, etc. In these applications, a piecewise linear model assumption is more preferred. In this paper, we propose a simple yet very effective framework to smooth images of piecewise linear model assumption using classical filters with the piecewise constant model assumption. Our method is capable of handling with gradient reversal artifacts caused by the piecewise constant model assumption. In addition, our method can further help accelerated methods, which need to quantize image intensity values into different bins, to achieve similar results that need a large number of bins using a much smaller number of bins. This can greatly reduce the computational cost. We apply our method to various classical filters with the piecewise constant model assumption. Experimental results of several applications show the effectiveness of the proposed method.