基于迭代思想的自适应Niblack算法改进

IMPROVEMENT OF ADAPTIVE NIBLACK ALGORITHM BASED ON ITERATIVE IDEA

针对文本及小物体图像识别时在光照不均匀工况下特征提取难度大、效率低问题,提出一种改进Niblack算法。利用迭代思想获取每个像素邻域的均值及标准差,使其与邻域窗口大小无关,避免了重复计算,降低了算法复杂度;在Sauvola算法基础上结合全局阈值Otsu算法及方差特性改进算法阈值,使其能自适应地调整阈值,逐点地进行图像二值化。为验证算法有效性,将其应用于文本、车牌及小物体图像二值化,与传统Otsu、Kittler、Niblack和Sauvola算法相比,算法获取的特征更清晰和准确,伪影和斑点等噪声更少,且与原Niblack算法相比,相同条件下其处理时间大幅度降低。

Aimed at the difficulty of extraction and low efficiency of text and small object image feature extraction under uneven illumination conditions,an improved Niblack algorithm is proposed.The iterative idea was used to obtain the mean and standard deviation of each pixel neighborhood,making it independent of the neighborhood window size,avoiding repeated calculations and reducing the algorithm complexity.Based on the Sauvola algorithm,the global threshold Otsu algorithm and the variance characteristics were combined to improve the algorithm threshold,so that it could adjust the threshold adaptively and perform image binarization point by point.In order to verify the effectiveness of the algorithm,it was applied to the binarization of text,license plate and small object images.Compared with the traditional Otsu,Kittler,Niblack,and Sauvola algorithms,the features obtained by the algorithm are clearer and more accurate,and the noises such as artifacts and spots are less.Compared with the original Niblack algorithm,its processing time is greatly reduced under the same conditions.