基于四元数离散Fourier变换的医学图像融合算法

针对医学图像融合中易忽略色彩过渡区域的信息,导致融合图像中存在色彩失真与纹理模糊的不足,本文提出了一种四元数离散Fourier变换的多通道彩色医学图像融合方案.首先,为了降低噪声干扰与颜色失真,将图像分割为一定大小的图像子块.其次,通过四元(Quaternion Numbers, QN)虚数映射像素的R,G,B分量,将每个块转换成QN表示.然后,将每个RGB块从空域变换到频域,使大部分图像信息集中到原点附近或少数几个区域,引入四元数离散Fourier变换(Quaternion Discrete Fourier Transform, QDFT),得到每个子块的QDFT系数.再测量和比较源图像的对比度值来得到融合QDFT系数.最后,在每个块上应用逆QDFT后,结合变换子块获得新融合图像.实验表明:与当前医学图像融合方案相比,本文方法具有更好的融合质量,其输出图像具有更加丰富的边缘与纹理,以及更高的对比度与分辨率.

Medical Image Fusion Algorithm Based on Quaternion Discrete Fourier Transforms

For color medical image fusion, it is easy to ignore the color transition area information, resulting in the lack of color distortion and fuzzy texture in fusion images, a multi-channel color medical image fusion scheme based on four Quaternion discrete Fourier transform was proposed. Firstly, in order to reduce noise and color distortion, the image was segmented into a certain size of image blocks. Secondly, each block was transformed to QN representation by mapping four pixels (Quaternion Numbers, QN) imaginary number to R, G and B components. Then, each RGB block was transformed from the space domain to the frequency domain, so that most of the image information was concentrated near the origin or a few regions, and the four elements number discrete Fourier transform (Quaternion Discrete Fourier Transform, QDFT) is introduced to obtain the QDFT coefficient of each sub block. Then the contrast value of the source image was measured and compared to get the fusion QDFT coefficient. Finally, inverse QDFT was applied to every block and combine all blocks to get the new fused image. The experiment shows that compared with the common medical image fusion scheme, the method has better edge and texture, the contrast and resolution were improved, the real-time and the robustness are strong, and it has good performance in the quantitative evaluation.