THESIS
2023
1 online resource (xi, 35 pages) : illustrations (some color)
Abstract
High-quality portrait photography has become an essential function in our daily lives.
However, due to the limited aperture and focal length of a smartphone camera, images
captured by a smartphone could not match the same level of bokeh effect by a digital
single-lens reflex camera. A common solution on a smartphone is to simulate out-of-focus
effects from an all-in-focus image, where the key is robust depth estimation and portrait
matting. To achieve this, we propose a multi-stage, multi-branch matting network to
estimate a strand-level portrait alpha mask, which is then used to refine the coarse depth
map from the pre-trained model. By integrating the input portrait image with the estimated
depth map and alpha mask, we propose a multi-plane image (MPI) representation
that includes mul...[
Read more ]
High-quality portrait photography has become an essential function in our daily lives.
However, due to the limited aperture and focal length of a smartphone camera, images
captured by a smartphone could not match the same level of bokeh effect by a digital
single-lens reflex camera. A common solution on a smartphone is to simulate out-of-focus
effects from an all-in-focus image, where the key is robust depth estimation and portrait
matting. To achieve this, we propose a multi-stage, multi-branch matting network to
estimate a strand-level portrait alpha mask, which is then used to refine the coarse depth
map from the pre-trained model. By integrating the input portrait image with the estimated
depth map and alpha mask, we propose a multi-plane image (MPI) representation
that includes multiple layers of disk-blurred images, where the kernel size is proportional
to the absolute depth distance to the focus layer. Then a depth-aware blurring process
is applied to enforce the bokeh effect. Besides, each MPI layer has an alpha channel controlling
the visibility according to the corresponding depth. Finally, an image with bokeh
is rendered by compositing all MPI layers. We conduct comprehensive experiments to
evaluate our method, which demonstrates that our method can generate more accurate
alpha masks and more realistic images with bokeh, compared to prior work.
Post a Comment