摘要
光学图层分解能够为物体识别提供有用信息,是图像内容理解和分析的基础。按照输入信息的不同可以将光学图层分解方法分为基于单幅图像和基于图集2类。对典型的光学图层分解方法进行论述和分析,包括交互式图层先验法、相对光滑度分解法、多重反射信息法、交互层叠法、反射变换法和稀疏性运动盲源法,阐述这些方法的原理和特点。结合软件测试平台验证和分析上述光学图层分解方法的优缺点及适用范围,并对光学图层分解方法的时效性、完备性和实用性进行展望。
Optical Layers Decomposition（OLD） can provide useful information for object recognition,and is the basis of image content understanding and analysis. According to the different input information, the OLD method can be divided into two kinds based on single image and atlas. This paper discusses and analyzes the typical OLD method, such as interactive layer prior method, relative smoothness decomposition method, multiple reflection information method, alternating stack method, reflection transformation method and sparse motion blind source method, explains the principles and characteristics of these methods. The software test platform is used to verify and analyze the advantages and disadvantages as well as the scope of application. The timelines,completeness and practicability of the OLD method are prospected.
出处
《计算机工程》
CAS
CSCD
北大核心
2017年第10期268-276,共9页
Computer Engineering
基金
中国博士后科学基金(30101166281)
中央高校基本科研业务费专项资金(20101166281
20103166281)
关键词
图层分解
病态问题
稀疏性
图层先验
成像模型
layer decomposition
ill-posed problem
sparsity
image prior
imaging model
