Abstract
Edge-aware image filter is an important tool in CG, CAD, and image processing applications. In this research, we propose a novel, fast, and high quality edge-aware color image filter. Our filter is based on a linear ridge regression approach with fast L1 Gaussian convolutions. Minimizing a cost function in our regression model is equivalent to apply a set of linear smoothing filters for image statistical moments. We employ a domain splitting algorithm for smoothing which possesses linear computational complexity with very accurate approximation. Thus, our filter is capable of producing high quality results quickly.
