针对现今学者提出的多种获取全景图像的图像镶嵌算法大都只是在图像发生平移条件下完成的局限性,提出一种广泛适用于图像发生任何空间变换(包括图像的平移、旋转、缩放和其它扭曲)的镶嵌算法。该算法核心是采用图像变形技术对发生空间变换的图像进行投影重建。具体过程是:首先由Harris算子提取的角点求得两幅图像问的变换矩阵,然后由此变换矩阵应用图像变形技术,逆向映射重采样插值发生空间变换的图像,最后与源图像合成图像。实验表明,该算法能达到良好的镶嵌效果。
As panorama image mosaic methods proposed by many scholars are completed mostly only under the translation conditions, a method of image mosaic is proposed which is widely suitable in the images that have any correlation in space (including image translation, revolving, zoom and other distortions). The core of the algorithm is using the technology of image morphing to perform the projection reconstruction of the image which has the correlation in space. The process is as follows: first, from the feature points ex- tracted by the Harris operator, the transformation matrix between two images is obtained; then, from this. transformation matrix, the technology of image morphing is adopoted, through inverse mapping, to resample and interpolate for the new image; finally, the interpolated image is synthesized with the source image. The experiments show that the method has good mosaic result.