融合Gabor纹理特征的观测场彩色图像均值偏移分割方法研究

A Method of Meteorological Observation Field Color Image Segmentation Using Mean Shift Combined with Gabor Texture Feature

摘要: 针对传统分割方法处理具有复杂性、多样性的室外彩色图像存在明显不足，本文提出一种融合Gabor纹理特征的室外彩色图像均值偏移分割方法。首先，采用Gabor滤波器组对图像进行纹理特征提取，将特征进行多方向融合降低特征维度。然后将纹理特征与图像像素的位置、颜色特征融合到均值偏移分割算法中，实现图像的区域分割。对比分水岭分割、传统均值偏移分割方法等，本方法能有效的控制过分割和欠分割的产生，能得到较好的分割效果。

Abstract: Due to shortage of traditional image segmentation methods dealing with complex and diverse outdoor color image, this paper puts forward to a Mean Shift segmentation method combined with Gabor texture feature. First of all, the paper extracts texture feature using Gabor filter and reduces the feature dimension by fusing multiple direction features. Then, mean space distance, color dis-tance and texture distance are calculated for region segmentation in images using Mean Shift clustering algorithm. Compared to watershed segmentation and classical Mean Shift clustering algorithm, this method can effectively control the generation of over-segmentation and owe seg-mentation and can get better segmentation effect.