基于长边检测的视频分割算法

针对前景和背景交界处颜色相似度较高时的图像分割问题,提出了基于长边检测的视频分割算法,首先建立包括颜色分量和对比度分量的能量函数,然后将基于边缘长度的边缘检测方法应用到基本模型的颜色模型中,利用长边检测的结果改进能量函数的颜色分量和对比度分量,最后使用图割算法,通过对能量函数求最小化得到最终的前景提取结果。实验结果表明,在前景和背景在交界处颜色相似度较高时,较其他算法具有明显的优势。

Video segmentation algorithm based on long-edge detection

Aiming at solving the problem in image segmentation when the image has high color similarity in the boundaries between foreground and background,a novel video segmentation algorithm based on long-edge detection is proposed.First the energy function is constructed with the color term and contrast term.Then a novel edge detection method based on edge length is applied to the result of color model in the basic model,and the color term and contrast term in the energy function are improved with the result of long edge detection.Finally the energy function is minimized by using the graph cutting method and the foreground extraction results are got.Experiments demonstrate that compared with other methods,our algorithm has obvious advantages in the case of high similarity in the boundaries between foreground and background.