摘要
归纳并分析了超像素算法和评价指标的最新研究成果及最新应用;对比了多种超像素算法的边缘召回率、欠分割错误率和紧凑度等评价指标,分析了各自的优势和不足。当前的超像素方法在精度和效率上都有较大的提高,应用领域不断增加,但仍难以满足特殊应用领域的超像素性能要求,需要研究稳健性和适应性更好的超像素算法。
The latest research results and applications of superpixel algorithms and evaluation indexes are summarized. Many superpixel methods are compared by using the evaluation indexes such as boundary recall, under-segmentation error rate, and compactness. The corresponding advantages and limitations are also analyzed. The experimental results show that the current superpixel methods are greatly superior to the previous methods in terms of accuracy and efficiency, and the applications of superpixel algorithms are growing constantly. However, it remains difficult to satisfy the requirements of the superpixel performances in some special applications. Therefore, it is necessary to develop the new methods that are more robust and have better adaptability.
作者
罗学刚
吕俊瑞
彭真明
Luo Xuegang;Lu Junrui;Peng Zhenming(School of Mathematics and Computer Science, Panzhihua University, Panzhihua, Sichuan 617000, China;School of Information and Com m unicat ion Engineering, University of Electronic Science and Technology of China ,Chengdu, Sichuan 610054, China)
出处
《激光与光电子学进展》
CSCD
北大核心
2019年第9期45-55,共11页
Laser & Optoelectronics Progress
基金
国家自然科学基金(61571096,61775030)
四川省教育厅科学研究项目(15ZB0425)
中国科学院光束控制重点实验室基金(2017LBC003)
关键词
图像处理
超像素
图像分割
评价标准
区域分割
image processing
superpixel
image segmentation
evaluation criterion
region segmentation
