目的 为解决目前多模图像时域联合滤波算法对图像细节信息保持较差的问题,提出一种多模图像交叉双域滤波算法。方法 在时域中使用交叉双边带滤波,通过多模图像边界上的信息互补保持边缘信息,然后对图像残量使用小波收缩算法恢复细节信息并叠加到时域滤波结果中。在此基础上构造时域和频域交替迭代并通过逐步递减缩小滤波核的范围获得最终滤波结果。结果 通过对多模医学图像和自然多模图像进行测试,相比目前联合滤波算法和单模双域算法,本文算法在峰值信噪比(PSNR)和视觉上都有较明显提高。结论 算法能够有效利用多模图像之间的互补信息,同时通过迭代有效抑制振铃负效应,将时域滤波及频率滤波的优势进行结合,使得滤波结果在保持高对比边缘的同时对图像细节也进行了较好的保留。并且该算法适用于所有含噪多模图像。

Objective This paper proposes a cross dual-domain filter to denoise multi-modal images and to address the poor performance of the time domain join filtering algorithm for multi-mode images. Method To maximize the use of complementary information at the edges of a multi-modal image, the proposed algorithm uses a cross bilateral filter in the time domain. Afterward, the residue is filtered using a wavelet shrinkage algorithm in the frequency domain to recover the detail texture information. The two domain results are then superimposed. The iteration between the time and frequency domains is constructed on this basis. The end result is progressively obtained by iteration over the adjusted parameters. Result Compared with the current noise-filtering methods that are based on either the single domain or single mode, the proposed algorithm shows great advantage in terms of visual quality, edge sharpness, and detail, and PSNR indicator. Conclusion The proposed algorithm can effectively use the complementary information between multi-mode images and combine the advantages of both the time and frequency domain methods. The negative ring bell effect can be effectively suppressed through the iteration.