基于小波变换极大模的多模医学图像融合

MULTIMODALITY MEDICAL IMAGE FUSION BASED ON WAVELET TRANSFORM MODULUS MAXIMA

多模态医学图像融合由于其对医学临床诊断的意义已引起广泛的关注,基于图像边缘特性的融合方法逐渐成为研究的重点。本文提出了一种基于离散二进小波变换的多尺度边缘检测和图像融合的方法,实现了特征级图像融合。不同于以往的从极大模值点直接重建图像,本文的算法利用极大模值点建立有效的融合规则,然后从融合的小波系数重构信号。融合图像的交互信息和峰值信噪比等检测指标表明此方法优于传统融合算法。

Multimodality medical image fusion is significant to clinical diagnosis, and it results in the extensive research on this field, most of which are focused on the methods based on image edge detection. Regarding this, a novel feature level approach to image fusion is proposed based on discrete dyadic wavelet transform for multi-scale image edge detection. Because it is very difficult and complicated to reconstruct image from wavelet modulus maxima, the modulus maxima is only used to construct effective fusion criteria in this algorithm and the fusion image is still reconstructed from fused wavelet coefficients. Mutual information and Peak Signal-to-Noise Ratio indicate that the method is better than traditional one based on wavelet transform. The experiment results also show that the algorithm has good performance both in speed and robustness.