Fuzzy based edge enhanced text detection algorithm using MSER
Fuzzy based edge enhanced text detection algorithm using MSER
复制标题
DOI：
10.1007/s10586-017-1448-5
复制
发表时间：
2017-12
期刊：
影响因子：
--
通讯作者：
A. Thilagavathy;A. Chilambuchelvan
中科院分区：
文献类型：
--
作者：
A. Thilagavathy;A. Chilambuchelvan
利用最大稳定极值区域（MSER）制作了一个创意系统，用于区分文本和图片。通过对具有较低分辨率的图像使用基本MSER，可能无法检测到小尺寸的文本。针对模糊图像，提出了一种基于模糊的边缘检测方法与MSER相结合的方法。边缘检测采用模糊推理系统。然后将水平和垂直投影轮廓沿着以及文本内容的几何质量连接到文本的单独的内容和非内容区域。然后通过使用边界框距离构建最小生成树来进行文本分组。该系统在ICDAR 2003数据集上进行了实验，显示了文本检测的潜在结果。
A creative system has been made with Maximally Stable Extremal Regions (MSER) for distinguishing apart the text from pictures. By using the basic MSER to images with lower resolution, the small sized text may not be detected. To deal with blurred images we put forward an approach that combines the fuzzy based edge detection method with MSER. The edge is detected using fuzzy inference system. Horizontal and vertical projection profiles along with geometric qualities of the text content are then connected to separate content and non-content locales of the text. Text grouping is then done by constructing the minimum spanning tree using bounding box distance. The proposed system is experimented on ICDAR 2003 dataset that shows potential outcomes on text detection.