结合假边缘提取和直方图分析的图像灰度变换

发布时间：2018-05-25 07:44

本文选题：刻划字符 + 直方图分析　；参考：《光学精密工程》2017年04期

【摘要】：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。
[Abstract]:The foreground of optical character image is usually composed of high gray pixel region and low gray pixel area, which leads to the traditional gradient-based edge extraction method such as Canny operator can not accurately obtain the contour of characters. In this paper, an image gray level transformation method based on false edge information and histogram analysis is proposed. The image foreground after transformation is only composed of low gray pixels. Firstly, the feature of Canny edge point is analyzed, then the false edge point is extracted, and then the image histogram analysis is carried out to determine the range of gray level transformation and the corresponding relation of gray level. The smooth degree constraint is applied to the boundary region on both sides of the false edge to determine the gray transformation parameters. Finally, the gray level transformation of the high gray pixel in the non-background region is carried out by the transformation parameter. The experimental results show that the character strokes are only composed of low gray pixels after gray level transformation, and the gray level changes at the original boundary are smooth enough to extract the complete and accurate character contours using gradient-based methods.
【作者单位】：山东建筑大学信息与电气工程学院;
【基金】：山东科技发展计划(No.2014GGX101050) 山东省自然科学基金(No.ZR2014FM016)
【分类号】：TP391.41

【相似文献】