采用精确抠图的单幅自然图像背景虚化合成方法

针对小光圈镜头下的单幅自然图像背景虚化问题，提出一种基于精确抠图的单幅自然图像背景虚化合成方法.首先，利用GrabCut算法对单幅自然图像的前景对象进行初始抠图；其次，采用基于局部二值模式算子的颜色恒常性算法提取图像纹理，并对初始抠图实现精确割图；然后，通过双边滤波虚化背景部分；最后，将精确分割后的前景部分与虚化后的背景部分合成.仿真结果表明：文中算法可以获得高质量的分割和视觉自然的背景虚化效果.

Synthesis Method for Background Blurring of Single Natural Image Using Accurate Matting

Aiming at the problem of background blurring of single natural image under small aperture lens, a synthesis method of background blurring of single natural image based on accurate matting is proposed. Firstly, the GrabCut algorithm is used to perform initial matting on the foreground object of single natural image. Secondly, the color constancy algorithm based on local binary patterns operator is used to extract the image texture, and the initial matting is precisely cut. Then, the background is virtualized by bilateral filter. Finally, the accurately segmented foreground part is combined with the virtualized background part. Simulation results show that the proposed algorithm can achieve high-quality segmentation and visually natural background blurring effects.