基于数据挖掘的图像特征分割技术

Image feature segmentation technology based on data mining

针对当前图像特征分割技术分割复杂多样图像时存在精度较低的问题,引入数据挖掘理念,研究了一种新的图像特征分割技术。使用K-means聚类算法进行聚类处理,依次使用柔化处理、中值滤波处理和锐化处理后,实现图像去噪。通过分析预处理图像获得白色线剖面图,根据数据挖掘确定图像中的目标颜色R/B值和背景颜色R/B值,在灰度共生矩阵中获得描述参量,对比对比度、熵、相关性和能量,实现纹理特征提取,由此完成复杂多样图像的高精度分割。分别对模拟图像和遥感图像进行分割实验,与传统分割结果进行对比,从定性和定量两个角度验证基于数据挖掘的图像特征分割技术的有效性,结果表明,基于数据挖掘的图像特征分割技术能够获得更全面的图像信息和纹理细节,从而更加精准地分割出同质区域。

As the current image feature segmentation technology has low precision when segmenting complex and diverse images,a new image feature segmentation technology is studied by introducing the idea of data mining.The K-means clustering algorithm is used for clustering processing,and then the image denoising is achieved by softening processing,median filtering processing and sharpening processing in sequence.A white line profile is obtained by analyzing the preprocessing image.The target color R/B value and the background color R/B value in the image are determined according to data mining technology.The description parameters are obtained in the gray-level co-occurrence matrix.The texture feature extraction is achieved by contrasting contrast,entropy,correlation and energy.On the basis of the above,the high-precision segmentation of complex and diverse images is completed.The segmentation experiments on simulated image and remote sensing image were performed.The results are compared with that of the traditional segmentation to verify the effectiveness of image feature segmentation technology based on data mining from both qualitative and quantitative perspectives.The results show that the image feature segmentation technology based on data mining can get more comprehensive image information and texture details,which facilitate the more accurate segmentation of the homogeneous region.