Lecture Node for Seam Carving
Zhengyang Gu Sun Yat-sen University guzhy3@mail2.sysu.edu.cn
Abstract
The lecture gave us some basic concepts about seam
carving, which are very intriguing. Therefore, I read the
papper[2] which rst introduce seam carving and some
other related pappers, in order to dig deeper into this al-
gorithm.
1. Introduction
Image retargeting is to change the size of pictures on a
website when the website is accessed by dierent divices
with various screen sizes. The simplest way is scaling and
cropping1. However, scaling may distort important contents
in pictures. For example, the cat1 becomes obviously dis-
torted after scaling. And cropping may remove important
contents at the periphery of pictures. For instance, some
of the tentacles near the edges of the picture1 are removed.
Therefore, content-aware image resizing approaches that
can maintain important contents when resizing pictures is
in great demand.
Seam carving is a content-aware image resizing appo-
rach. The basic idea of it is to remove the most unimpor-
tant pixels in a seam which is intuitively a line. Therefore, it
can eliminate redundance and keep consistency of pictures
at the same time. For example, the picture2 is resized by
seam carving, while the cat is not distorted and the tentacles
are not removed.
2. Background
There are some other content-aware image resizing meth-
ods mentioned by the paper[2]. For example, use top down
methods like face detectors[10] or bottom down methods
like visual saliency methods[7] to build saliency map. And
then based on the saliency map, use intelligent cropping
methods to resize the image[9]. Another paper[4] pays at-
tention to users’ perception on the adapted images. Their
method automatically detect and transmit the most impor-
tant part of the image to devices.
And the paper[2] also mentioned other areas where seams
are used. For instance, the paper[1] nds the best seams
to combine parts of a set of pictures into a single compos-
ite picture. And the computation of seams can be done by
(a) Original
(b) Scaling (c) Cropping
Figure 1: The original picture shows a cat with some tenta-
cles. The picture is scaled and cropped.
Figure 2: Seam carving
other ways, including Dijkstra’s shortest path algorithm[6]
and graph cuts[3].
1
下载后可阅读完整内容，剩余8页未读，立即下载
kdbshi
- 粉丝: 168
- 资源: 298
我的内容管理
展开
最新资源
- 十种常见电感线圈电感量计算公式详解
- 军用车辆：CAN总线的集成与优势
- CAN总线在汽车智能换档系统中的作用与实现
- CAN总线数据超载问题及解决策略
- 汽车车身系统CAN总线设计与应用
- SAP企业需求深度剖析：财务会计与供应链的关键流程与改进策略
- CAN总线在发动机电控系统中的通信设计实践
- Spring与iBATIS整合：快速开发与比较分析
- CAN总线驱动的整车管理系统硬件设计详解
- CAN总线通讯智能节点设计与实现
- DSP实现电动汽车CAN总线通讯技术
- CAN协议网关设计：自动位速率检测与互连
- Xcode免证书调试iPad程序开发指南
- 分布式数据库查询优化算法探讨
- Win7安装VC++6.0完全指南：解决兼容性与Office冲突
- MFC实现学生信息管理系统：登录与数据库操作
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
