结合假边缘提取和直方图分析的图像灰度变换

光学刻划字符图像的前景通常由高灰度像素区域和低灰度像素区域组成,导致使用传统的基于梯度的边缘提取方法如Canny算子不能准确地获得字符的轮廓,为此本文提出了一种利用假边缘信息结合直方图分析的图像灰度变换方法,变换后的图像前景仅由低灰度像素构成。首先对Canny边缘点进行特征分析,提取其假边缘点;然后进行图像直方图分析,确定灰度变换范围以及灰度对应关系,并对假边缘两侧的边界区域施加光滑度约束来确定灰度变换参数;最后以此变换参数对非背景区域的高灰度像素进行灰度变换。实验证明,灰度变换后字符笔画仅由低灰度像素构成,原边界处的灰度变化足够光滑,可以使用基于梯度的方法来提取完整而准确的字符轮廓。

detection with histogram analysis

As optical carving character image foreground often consists of high gray-scale pixel region and low gray-scale pixel region, it is impossible to gain character profile correctly with traditional edge extraction method on the basis of gradient, such as Canny operator, and therefore, a kind of image gray-scale transformation method combining false edge information and histogram analysis was introduced.The image foreground after transformation consists of low gray-scale pixel merely.Firstly, feature of Canny edge point was analyzed to extract false edge points.Then histogram analysis of image was performed to determine gray-scale transformation scope and corresponding relationship.Then, smoothness constraints was imposed on boundary region at both sides of false edge to determine gray scale transformation parameters.Finally, on the basis of transformation parameters, gray-scale transformation was carried out on high gray-scale pixel not belonging to background region.Experiment result shows that character strokes after gray-scale transformation consist of low gray-scale pixel merely.Gray-scale transformation at the original boundary is smooth enough and method on the basis of gradient can be used to extract complete and correct character profile.