Double-Guided Filtering: Image Smoothing with Structure and Texture Guidance

Abstract

Image smoothing is a fundamental technology which aims to preserve image structure and remove insignificant texture. Balancing the trade-off between preserving structure and suppressing texture, however, is not a trivial task. This is because existing methods rely on only one guidance to infer structure or texture and assume the other is dependent. However, in many cases, textures are composed of repetitive structures and difficult to be distinguished by only one guidance. In this paper, we aim to better solve the trade-off by applying two independent guidance for structure and texture. Specifically, we adopt semantic edge detection as structure guidance, and texture decomposition as texture guidance. Based on this, we propose a kernel-based image smoothing method called the double-guided filter (DGF). In the paper, for the first time, we introduce the concept of texture guidance, and DGF, the first kernel- based method that leverages structure and texture guidance at the same time to be both 'structure- aware' and 'texture-aware'. We present a number of experiments to show the effectiveness of the proposed filter.