Interactive 2D and Volume Image Segmentation Using Level Sets of Probabilities

彬彬

1 Introduction

In this technical sketch, we adopt the level set method for image

segmentation that integrates region statistics and edge responses. It

is well-known that a serious limitation of existing level set algorithms

for image segmentation is that the final result is sensitive to

the location of the initialization. This is because level set evolution

is typically driven by forces computed from local image data.

We overcome this problem by adopting a novel level set function

based on foreground probabilities, and further integrating the level

set method with a probabilistic pixel classifier [Liu and Yu 2012].

Since an accurate classifier does not exist at the beginning, the

segmentation framework is based on the expectation-maximization

(EM) algorithm. In summary, the motivations for our method based

on level sets of probabilities are manifold.

• It is possible to achieve a good initialization of our new level

set function with the probabilistic pixel classifier. An EM-based

algorithm is capable of improving the performance of both the perpixel

classifier and the level set method over multiple passes, further

making final object segmentation less sensitive to initialization.

• The zero level set can represent the boundary of an object with an

arbitrary topology. It is also very convenient to evolve the topology

of the zero level set during the solution process. Thus, the level

set method can be effectively used for extracting a foreground layer

with fragmented appearances, such as leaves and cells.

• It is possible to make the zero level set snap to relatively distant

salient edges by devising an effective force based on the location of

these edges. We achieve this goal using a specially designed edge

distance field based on Canny edge detection.

C.R.CAN

有意思！学习了！

C.R.CAN

都闪开，介个帖子，偶来顶