多方法融合的类圆形堆积物分割

Segmentation for Accumulation of Qusic-circular Object with Multiplex Methods

针对类圆形堆积物图像的前景和背景在色彩或灰度上相近,难以用传统阈值分割等算法进行有效分割的问题,提出一种多方法融合的类圆形堆积物图像区域分割策略。对图像进行滤波等一系列预分割处理后进行投影得到目标图像的外接矩形区域,以排除噪声的干扰,在区域内采用改进的灰度共生矩阵方法进行粗分割,以解决窗口大小与分割精度的问题,采用二维OTSU阈值分割方法对粗分割结果进行量化。实验结果表明,该方法得到的区域分割结果边缘清晰、准确度高。

This paper proposes a segmentation method for accumulation of qusic-circular object with multiplex methods, as the image＇s foreground and background is simple at color or gray level. It uses a series method such as filtering to get a preprocess result, and does spot projection to get circum-rectangle of the accumulation object in order to get rid of the noise＇s interference. It uses improved gray matrixes to reserve the contradiction of window＇s size and the precision of segmentation. 2D OTSU threshold is used to get a precise segmentation result. Experimental results show that the area segmented by the method is clear and with high veracity.