We present a new local window based image processing framework, which is particularly effective on edge-preserving and texture-removing. This seemingly contradictive effect is achieved by combining the traditional full window filtering strategy (FWF), which is good at removing noise, and the recently proposed side window filtering (SWF) strategy, which is good at preserving edges, so the new framework is called combined window filtering (CWF). By using window inherent variation method, we can easily distinguish the edges of structures from the texture. For the pixels on edges, SWF is used to preserve them and for the pixels on texture, FWF with multiple scales is used to remove them. This technique is surprisingly simple yet very effective in practice. We show that many traditional linear and nonlinear filters can be easily implemented under CWF framework. Extensive analysis and experiments show that implementing the CWF principle can significantly improve their edge-preserving and texture-removing capabilities and achieve state of the art performances in applications.

中文翻译：

---

组合窗口过滤及其应用

我们提出了一种新的基于局部窗口的图像处理框架，它在边缘保留和纹理去除方面特别有效。这种看似矛盾的效果是通过结合传统的擅长去除噪声的全窗滤波策略（FWF）和最近提出的擅长保留边缘的侧窗滤波（SWF）策略来实现的，因此新框架被称为组合窗口过滤 (CWF)。通过使用窗口固有变化方法，我们可以很容易地将结构的边缘与纹理区分开来。对于边缘的像素，使用 SWF 来保留它们，对于纹理上的像素，使用具有多个尺度的 FWF 来移除它们。这种技术非常简单，但在实践中非常有效。我们展示了许多传统的线性和非线性滤波器可以在 CWF 框架下轻松实现。大量的分析和实验表明，实施 CWF 原理可以显着提高其边缘保留和纹理去除能力，并在应用中实现最先进的性能。