基于图像分块的局部阈值二值化方法

发布时间：2018-03-21 21:49

本文选题：图像二值化　切入点：图像分块　出处：《计算机应用》2017年03期 　论文类型：期刊论文

【摘要】：针对目前局部阈值二值化结果存在目标虚假或断裂的缺陷,提出了一种基于图像分块的局部阈值二值化方法。首先,将图像分成若干子块并分析每个子块像素灰度变化情况;接着,取一定大小的局部窗口在图像中移动,比较该局部窗口内与包含窗口自身且比窗口更大区域内的像素灰度变化情况,更大区域由窗口模板当前覆盖的所有子块组成,以此判断窗口内是否为灰度变化平坦(或剧烈)区域;最后,根据不同的区域,给出具体的二值化方案。利用7种不同算法对4种不同类型的4组图像进行了二值化实验。实验结果表明该算法在屏蔽背景噪声和保留目标细节方面表现最优,特别地通过对车牌图像的二值化结果进行定量分析后发现该算法能够得到最高召回率和准确率。
[Abstract]:Aiming at the defect of false or broken target in the current local threshold binarization results, a local threshold binarization method based on image partitioning is proposed. Firstly, the image is divided into several sub-blocks and the gray level changes of each sub-block pixel are analyzed. Then, a local window of a certain size is taken to move in the image to compare the changes of pixel grayscale between the local window and the region containing the window itself and in a larger area than the window. The larger region is composed of all the sub-blocks currently covered by the window template. To determine whether the window is a flat (or violent) region of grayscale change; finally, depending on the region, A specific binarization scheme is presented. Four groups of images of four different types are binarized by using seven different algorithms. The experimental results show that the algorithm performs best in shielding background noise and preserving the details of the target. In particular, through the quantitative analysis of the binarization results of license plate images, it is found that the algorithm can obtain the highest recall rate and accuracy.
【作者单位】：中国药科大学理学院;
【基金】：国家自然科学基金资助项目(61501522)~~
【分类号】：TP391.41

【相似文献】