- 4
- 0
- 约5.32万字
- 约 66页
- 2021-10-11 发布于天津
- 举报
毕业设计(论文)说明书
题目: 复杂背景图像中文本定位算法设计
摘要 随着多媒体技术的飞速发展, 复杂背景图像中的文本定位研究不仅丰富了图 像处理理论,而且在诸如In ternet环境下的图像检索、交通管理中的车牌识别等 具有重大的价值。 复杂背景文本定位是一个具有较大难度性的研究课题, 原因是 文本图像的背景非常复杂, 图像大多在室外拍摄, 光照条件变化较大, 其中不同 文字的颜色、亮度、字体、大小、间距、对比度、排列方向和背景纹理等有很大 差别。要提取具有复杂背景的文本, 首先要找到包含文本的区域, 然后才能利用 文本识别模块进行识别。 本文综述了现有的主要文本定位方法, 分析了其中的优 缺点,实现了一种基于边缘检测和支持向量机的图像文本定位方法。 其中,基于 边缘检测的文本定位主要由金字塔分解、基于改进 Canny算子的边缘检测、边缘 提取和二值化、 连通区域分析、 以及文本区域鉴定与合并几部分组成。 首先运用 改进的Canny边缘检测算法检测出文本边缘,然后对检测结果进行连通区域分析、 文本区域鉴定与合并得到候选的文本区域。 进一步,通过将定位出的候选文本区 域运用支持向量机的分类器训练的方法来提高文本定位的准确性。实验结果表 明,该文本定位方法不但可以较准确的定位出相应的文本区域, 而且具有一定的 意义和较大的实用价值。
关键词:文本定位;边缘检测;特征提取;支持向量机
ABSTRACT
With the development of the multimedia technology, the study of locating texts under complicated background has not only enriched image processing theoretically, but also has enormous value in practical application. For example, the image retrieval under Internet environment and the discernment of the plate number in traffic administration. The location and extraction of text from complex background is an important research problem in the computer vision . The variation of the text in
terms of characters font, size, style, orientation alignment, texture color and complex background makes the problem of text localization very difficult. The scene content is unconstrained and maybe both indoor and outdoor scenes under any lighting or contrast conditions .
To extract complex background text, text areas should be located first .Current text location methods ale researched in this paper, and the advantage and disadvantage of them are analyzed .Then text location
method based on edge detectionand supportvector machines is
method based on edge detection
and support
vector machines is
implemented.
Edge detectionbased text locationmethod is composed by Pyramiddecomposition,improved Canny algorithm-basededge detection,edgeabstractingand binary,connected component analysis,text region
Edge detection
based text location
method is composed by Pyramid
decomposition,
improved Canny a
原创力文档

文档评论(0)