图像/视频字符区域定位算法的实现

Realization of text detection and location in image/Video

对图像/视频中的文字检测方法是基于字符的纹理特征进行处理的,首先将图像进行 DCT 变换,因为变换的系数分布代表了不同方向的能量,对变换的系数进行能量计算得到一个比较粗略的选区,然后进行形态学操作,滤掉一些不符合字符特征的区域,从而将图像中的字符区域提取出来,用 matlab 进行了测试,测试中发现,对一些复杂的背景下的字符还存在着漏判和误判现象。

It is important to detect and locate the text in image/video,arithmetic is based on the character of text texture.First we process a DCT transformation on the image/video,the DCT coefficient can represent the energy of different direction.Second we calculate the energy to get a rough region,then make morphological filter.So we can recognize the text region from background.We test it with Matlab.Find it can make also wrong judgement when in complex background.