摘要
由于基于小波变换的图像融合方法仅在水平、垂直、对角线三个方向对图像的高频信息进行分解,易造成图像轮廓的不连续性,而基于非下采样轮廓波变换(NSCT)的融合方法能够对图像中的高频信息进行多方向分解。本文基于NSCT对多模态医学图像进行融合,以区域能量平均加权法作为NSCT变换后低频子带的融合规则,区域能量中心加权作为高频内层带通子带的系数融合规则,外层带通子带则采用区域能量最大的系数融合规则。通过对精确配准的头部PET、CT、MR图像的融合实验,借助清晰度、信息熵、联合熵等指标进行客观评价,‘prewitt’算子提取融合图像的边缘信息进行主观评价,验证该算法在视觉效果、信息含量、实时性和长轮廓表达上的优势。相较于其他文献中提出的算法,本文算法得到的融合结果信息丰富程度提升约7%、清晰度提升约31.7%、程序运行时间缩短一半,应用前景可观。
Image fusion method based on wavelet transform just decomposes the high frequency information in horizontal,vertical and diagonal directions, easily resulting in the discontinuity of image contour, while the fusion method based on non-subsampled contourlet transform（NSCT） decomposes the high frequency information in multi-directions. In the paper,the coefficients of the low frequency sub-band were combined by the average weighting regional energy rule; the innermost coefficients of the high frequency band were combined by the weighting regional energy rule; the outermost coefficients of the high frequency bands were combined by the maximal regional energy rule. Based on the experiment of fusing the registered positron emission tomography, CT and MR images of head, the clarity, information entropy and joint entropy were applied for objective evaluation evaluate. And the edge information of the fused image was extracted by ＇prewitt＇operator to make subjective assessment, proving the advantages of the proposed algorithm in visual effect, information content, real-time and the long contour expression. Compared with algorithms in two published papers, the image fused by the algorithm proposed in the paper has about 7% increase in information content and 31.7% improvement in clarity, and reduces nearly a half of the program running time, showing fantastic potential applications.
出处
《中国医学物理学杂志》
CSCD
2016年第5期445-450,共6页
Chinese Journal of Medical Physics
基金
国家自然科学青年基金(61201067)
上海市教委科研创新项目(13YZ069)
关键词
图像融合
小波变换
非下采样轮廓波变换
区域能量
平均梯度
image fusion
wavelet transform
non-subsampled contourlet transform
regional energy
average gradient
