硕士论文-IN1feng.doc

摘 要 解译有效利用遥感数据，高分辨率遥感图像包含丰富的颜色、纹理等信息，图像本身也含有大量噪声，因此如何较好地移除噪声，并合理利用图像所包含的各类信息图像准确分割高分辨率遥感图像算法图像滤波、颜色量化和空间分割图像中区域边界较模糊分割区域内不同阴影过分割现象。利用局部同质矩阵校正局部值，准确反映。减弱或消除传统算法的过分割现象稳定描述图像纹理特征算子具有相纹理信息的颜色类图合并。采用分析法、优度实验法中的评价测度，在算法中值和均值计算的基础上给出了值和均值评价方法评价。 实验结论表明本文提出的改进算法克服算法。关键词: 高分辨率遥感图像，算法，局部同质，算子AbstractIn order to effectively analyze and interpret the remote sensing data, the remote sensing image processing technology has become hotspot. The remote sensing image segmentation is one of key factors, which decide the success of remote sensing image analyze and calculation. Because only if the better segmentation effect is obtained, the follow-up works can be get better effects, such as information extraction and target recognition et al. The background of this paper is high resolution remote sensing image. As these remote sensing images contain a wealth of color, texture and a lot of noise et al, the key point of the research is how to effectively remove noise, smooth the image and reasonably use various information to segment the image. Firstly, the image filtering, color space quantization and spatial segmentation can be completely simultaneously by use the JSEG algorithm, which directly utilize color information. Then, In order to solve the limitation case of the JSEG, such as over segmentation, the improved JSEG algorithm is proposed in this paper. For obtaining the better effect of segmentation, the local J value is corrected by local homogeneous matrix, which gives reasonably description for color homogeneity in region and the boundary of different region. In addition, for weakening or eliminating the over segmentation, the class map is merged by LBP/C operator. It gives stable description for the texture of image. The finally, the evaluation of image segmentation is analyzed and summarized. In addition, the evaluation method of J value and J average value is proposed based on calculating the J value and J average value in JSEG algorithm. The merit is reflected by objective