- 1、本文档共5页,可阅读全部内容。
- 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
- 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载。
- 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
- 5、该文档为VIP文档,如果想要下载,成为VIP会员后,下载免费。
- 6、成为VIP后,下载本文档将扣除1次下载权益。下载后,不支持退款、换文档。如有疑问请联系我们。
- 7、成为VIP后,您将拥有八大权益,权益包括:VIP文档下载权益、阅读免打扰、文档格式转换、高级专利检索、专属身份标志、高级客服、多端互通、版权登记。
- 8、VIP文档为合作方或网友上传,每下载1次, 网站将根据用户上传文档的质量评分、类型等,对文档贡献者给予高额补贴、流量扶持。如果你也想贡献VIP文档。上传文档
查看更多
Isolated Character Recognition by Searching Feature Points
Isolated Character Recognition by Searching Feature Points
Abstract
Conventional segmentation technique cannot extract a
isolated character and touching characters. In this paper,
we propose a novel character recognition method which ex-
ecutes segmentation and recognition simultaneously. This
method enable to extract such difficult characters to extract.
The effectiveness of the proposed method is confirmed by
experiments.
1. Introduction
Accuracy of character recognition for segmented char-
acter image is enough for practical use. However, con-
ventional segmentation method can be applied only well-
defined problems requiring that the character images are ar-
ranged on a line, completely separated from other charac-
ters, and so on. Therefore, isolated character which is apart
from lines of characters, and touching characters, which is
parts of more than two characters are connected, are hard to
extract. In this paper, we propose a novel character recogni-
tion method which executes segmentation and recognition
simultaneously. The proposed method allow segmentation
of a isolated character, touching characters and so on.
2. Preparation
2.1. Input image and reference image
“Input image” is a relatively larger image including tar-
get characters. “Reference image” is a relatively smaller
image among 62 characters of alphabets and numerals.
2.2. Histogram of edge direction
At first, the edge direction and the edge intensity of each
pixel is calculated from horizontal and vertical intensity by
the Sobel filter. Then, the edge direction is quantized into n
bins. Each bin has θ = 2π
n
radian. To eliminate the effect
of noise, a pixel whose intensity is less than a threshold is
regarded as a no edge pixel. The n bins do not include no
(a) Corners.
Up
Left Right
Down
(b) Curves.
Figure 1. Features.
edge pixels. The histogram are normalized so that the sum
of the histogram values to be 1. This normalized histogram
is a feature vector of n dimensions.
2.3. Feature point
Corn
您可能关注的文档
- imm5645 IMM5257E 加拿大签证 填表攻略.pdf.pdf
- Impact of anxiety profiles on cognitive performance in BALBc.pdf
- Impact analysis of nanoparticle motion me__ chanisms on the thermal conductivity of n__ anofluids.pdf
- IMMETECH INDUSTRIAL CO.,LTD_EN.pdf
- Impact of nitrogen fertilization and soil tillage on arbuscular mycorrhizal fungal communities.pdf
- Impact of nitrogen reduction measures on nitrogen.pdf
- Impact Testing and Simulation of a Crashworthy Composite Fuselage Section.pdf
- Implementing Set-Oriented Production Rules as an Extension to Starburst.pdf
- Implications of the Quark Mass Hierarchy on Flavor Mixings.pdf
- image rejection measure.pdf
文档评论(0)