一种基于方向性信息测度和IHS变换的图像融合新方法

提出了一种基于图像方向性信息测度和IHS变换的图像融合新方法。该方法首先根据不同邻域尺度下图像方向性信息测度的变化规律．检测出SAR图像象素中的边缘点和纹理点。然后，对TM图像的5、4、3波段(即红、绿、蓝波段)进行IHS变换得到亮度图像Ⅰ。进而根据SAR图像象素分类情况将其和Ⅰ图像进行线性加权融合，其中边界象素和纹理象素的权值比平滑象素大。最后通过IHS反变换，得到完整的融合结果。实验结果表明，与传统的IHS融合方法和基于直方图匹配的IHS融合方法相比，该新方法在保持TM光谱信息和抗噪声能力方面具有明显优势。

A New and Better Image Fusion Method Based on Orientation Information Measurement and IHS Transform

Existing fusion methods suffer, in our opinion, from two shortcomings: (1) much loss of spectrum information; (2) insufficient resistance to noise. We propose a new fusion method to overcome these two shortcomings. First, we use orientation information measurement theory to classify pixels in the SAR (Synthetic Aperture Radar) image into three classes: smooth, edge and texture. Then we use IHS (intensity, hue, saturation) transform to extract I (intensity) image in three wavebands: TM5 (red), TM4 (green), TM3 (blue). Next, we fuse intensity image with SAR image by different weighting strategies, whose weight for edge and texture pixels is higher than that for smooth pixels. Finally, we obtain the whole fusion image by inverse IHS transform. Compared with other algorithms, the experimental results of the proposed algorithm are much better in resisting noises. We also compare the correlations of experimental results of three fusion methods with original TM image. They show that TM5 correlations of two existing methods and our method are: traditional IHS method 0.391 3, IHS method based on histograms 0.468 4, and our method 0.805 4. The experimental results for TM4 and TM3 correlations are similar. The results above show that our method is better than existing methods in preserving spectrum information.