基于单目深度估计方法的图像分层虚化技术

Hierarchical image blurring technology based on monocular depth estimation method

  • 摘要: 实现了一种基于单目深度估计方法的图像分层虚化技术,可使前景清晰背景模糊,离镜头越远虚化效果越明显:首先对图像等比例缩小并且对图像中的物体做边缘检测,并利用线性追踪生成线条扫描信息图,利用双边滤波器对初始深度图做平滑处理得到深度图,依据深度信息选择阈值将图像分层,在分层的基础上对图像背景做高斯模糊处理,最终得到图像分层虚化效果,增强图像的表现力和艺术效果. 本技术的实现效果较为理想,能达到良好的图像分层虚化效果。

     

    Abstract: A method of layered image virtualization based on monocular depth estimation is proposed, which can make the foreground clear and the background blurred. When it is farther away from the lens, then it is more obvious for the effect of virtualization. First of all, the image is reduced in proportion, and the edges of objects in the image are detected, and generating line scan information map using linear tracing . A depth map is obtained by smoothing the initial depth map with a bilateral filter. The image background is layered according to the depth information of image. The Gauss fuzzy processing is done in the layered to obtain, the final image layered virtual effect, which enhances the image expression and artistic effect. The effect of this technology is ideal, and it can achieve a good effect of layered image blurring.

     

/

返回文章
返回