Cascaded Segmented Matting Network for Human Matting

Human matting, high quality extraction of humans from natural images, is crucial for a wide variety of applications such as virtual reality, augmented reality, entertainment and so on. Since the matting problem is an ill-posed problem, most previous methods rely on extra user inputs such as trimap o...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Bo Liu, Haipeng Jing, Guangzhi Qu, Hans W. Guesgen
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	Human matting semantic segmentation salient object detection Electrical engineering. Electronics. Nuclear engineering TK1-9971
Acceso en línea:	https://doaj.org/article/dec425f05e4e41328fbaf49abdaa4b2d
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:dec425f05e4e41328fbaf49abdaa4b2d
record_format	dspace
spelling	oai:doaj.org-article:dec425f05e4e41328fbaf49abdaa4b2d2021-12-02T00:00:24ZCascaded Segmented Matting Network for Human Matting2169-353610.1109/ACCESS.2021.3125356https://doaj.org/article/dec425f05e4e41328fbaf49abdaa4b2d2021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9600805/https://doaj.org/toc/2169-3536Human matting, high quality extraction of humans from natural images, is crucial for a wide variety of applications such as virtual reality, augmented reality, entertainment and so on. Since the matting problem is an ill-posed problem, most previous methods rely on extra user inputs such as trimap or scribbles as guidance to estimate alpha value for the pixels that are in the unknown region of the trimap. This phenomenon makes it difficult to be applied to large scale data. In order to solve these problems, we studied the unique role of semantics and details in image matting, and decomposed the matting task into two sub-tasks: trimap segmentation based on high-level semantic information and alpha regression based on low-level detailed information. Specifically, we proposed a novel Cascaded Segmented Matting Network (CSMNet), which uses a shared encoder and two separate decoders to learn these two tasks in a collaborative way to achieve the end-to-end human image matting. In addition, we established a large-scale dataset with 14,000 fine-labeled human matting images. A background dataset is also built to simulate real pictures. Comprehensive empirical studies on above datasets demonstrate that CSMNet could produce a stable and accurate alpha matte without the input of trimap and achieve an evaluation value that is comparable to the algorithm that requires trimap.Bo LiuHaipeng JingGuangzhi QuHans W. GuesgenIEEEarticleHuman mattingsemantic segmentationsalient object detectionElectrical engineering. Electronics. Nuclear engineeringTK1-9971ENIEEE Access, Vol 9, Pp 157182-157191 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Human matting semantic segmentation salient object detection Electrical engineering. Electronics. Nuclear engineering TK1-9971
spellingShingle	Human matting semantic segmentation salient object detection Electrical engineering. Electronics. Nuclear engineering TK1-9971 Bo Liu Haipeng Jing Guangzhi Qu Hans W. Guesgen Cascaded Segmented Matting Network for Human Matting
description	Human matting, high quality extraction of humans from natural images, is crucial for a wide variety of applications such as virtual reality, augmented reality, entertainment and so on. Since the matting problem is an ill-posed problem, most previous methods rely on extra user inputs such as trimap or scribbles as guidance to estimate alpha value for the pixels that are in the unknown region of the trimap. This phenomenon makes it difficult to be applied to large scale data. In order to solve these problems, we studied the unique role of semantics and details in image matting, and decomposed the matting task into two sub-tasks: trimap segmentation based on high-level semantic information and alpha regression based on low-level detailed information. Specifically, we proposed a novel Cascaded Segmented Matting Network (CSMNet), which uses a shared encoder and two separate decoders to learn these two tasks in a collaborative way to achieve the end-to-end human image matting. In addition, we established a large-scale dataset with 14,000 fine-labeled human matting images. A background dataset is also built to simulate real pictures. Comprehensive empirical studies on above datasets demonstrate that CSMNet could produce a stable and accurate alpha matte without the input of trimap and achieve an evaluation value that is comparable to the algorithm that requires trimap.
format	article
author	Bo Liu Haipeng Jing Guangzhi Qu Hans W. Guesgen
author_facet	Bo Liu Haipeng Jing Guangzhi Qu Hans W. Guesgen
author_sort	Bo Liu
title	Cascaded Segmented Matting Network for Human Matting
title_short	Cascaded Segmented Matting Network for Human Matting
title_full	Cascaded Segmented Matting Network for Human Matting
title_fullStr	Cascaded Segmented Matting Network for Human Matting
title_full_unstemmed	Cascaded Segmented Matting Network for Human Matting
title_sort	cascaded segmented matting network for human matting
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/dec425f05e4e41328fbaf49abdaa4b2d
work_keys_str_mv	AT boliu cascadedsegmentedmattingnetworkforhumanmatting AT haipengjing cascadedsegmentedmattingnetworkforhumanmatting AT guangzhiqu cascadedsegmentedmattingnetworkforhumanmatting AT hanswguesgen cascadedsegmentedmattingnetworkforhumanmatting
_version_	1718404017160192000

Cascaded Segmented Matting Network for Human Matting

Ejemplares similares