改进的Welsh颜色迁移算法在GF-2图像融合中的应用

Application of improved Welsh's color transfer algorithm to GF-2 image fusion

图像融合过程中需要对遥感图像进行几何纠正、配准等预处理,在国土资源调查研究工作中是一项繁重工作。因此,提出利用改进的Welsh颜色迁移算法实现图像融合,以湖北省仙桃市GF-2高空间分辨率全色图像(简称高分图像)为研究数据进行实验,并将迁移结果与Gram-Schmidt(GS)和主成分(principal components,PC)融合结果进行定性和定量对比。结果表明,基于Welsh颜色迁移的融合算法在图像光谱和纹理的保真度上都比GS和PC融合算法有更好的效果;特别是在地物复杂区域,颜色迁移算法的纹理结构相似度远优于GS和PC融合算法。同时,颜色迁移的算法可以应用于不同区域的多光谱和全色波段图像融合,而无需先对图像进行几何纠正、配准等预处理。

During the processing,image fusion,calibration,registration and other pre-processing steps for images are onerous tasks in land resources investigation. In this paper,the improved Welsh color transfer was proposed for image fusion. The study area was Xiantao City in Hubei Province and the study images were GF-2 high spatial resolution image data. The result of color-based image fusion algorithm was qualitatively and quantitatively compared with that of Gram-Schmidt（ GS） and principal components（ PC） fusion methods. The results show that the improved Welsh color transfer has good performance on preserving color fidelity and texture similarity.Especially in complex land-surface areas,the texture similarity of result image based on improved color transfer method is much better than that based on GS and PC fusion algorithm. The color transfer method is also applied to images fusion about different areas without calibration and registration for images.