引导滤波与视觉感知结合的自然场景轮廓提取

为更好地解决复杂背景下非结构化自然场景对象的轮廓提取问题,提出了一种引导滤波与视觉感知机制相结合的方法。通过改进的加权引导滤波方法对输入图像进行预处理,在去除图像中噪声的同时,保留和突出图像中物体的轮廓信息;之后采用基于非经典感受野视觉机制的背景抑制和轮廓增强算法进行处理。给出了完整的处理流程,阐述了各部分算法的实现步骤。通过最后定性和定量的实验对比,所提出的算法在效果上优于以往的同类算法,为进一步自然场景轮廓提取的实用化改进和视觉感受野仿生图像处理提供了一种思路。

Natural scene contour detection by combining guided filter and visual perception

To effectively deal with the problem of contour detection in non-structuring natural scene images under complex background, an algorithm combining guided filter and visual perception is proposed. Images are pre-processed through the improved weighted guided filter algorithm to remove the noise, preserve and enhance contour information of the object; then the images are processed through texture inhibition and contour strengthen algorithm inspired by visual perception mechanism of non-classical receptive fields. Complete implementation procedure and processing steps are presented. The qualitative and quantitative experimental results demonstrate that the proposed contour detection method outperforms other methods that deploy inhibition. The method provides a novel approach for practical improvement of natural scene contour extraction and visual imitation image processing.