基于Retinex和SURF的医学图像配准与拼接

Medical image registration and mosaic based on Retinex and SURF algorithms

针对医疗成像设备拍摄范围受限,以及医学图片受电子噪声和射线本身易散射、照度不均等复杂因素影响干扰医生诊断病情的问题,提出了基于Retinex和SURF的医学图像配准与拼接算法。算法首先通过三边滤波器改进多尺度Retinex方法增强图像,利用相位相关互功率谱快速估计重叠区域,减少配准范围;接着通过SURF局部特征提取算法提取特征点并精确提纯,建立图像变换矩阵实现拼接;最后针对接缝明显的现象,利用提升小波变换多分辨率融合算法进一步对图像进行融合处理,使得图像接缝过渡平滑自然。实验及质量评价标准的结果证实,新方法处理的医学图像配准与拼接对复杂环境有较高的鲁棒性。

According to the problems of the limited range of the camera, complex factors with the electronic noise, X-ray scattering and uneven illumination, they often disturb the doctor＇s diagnosis. This paper developed a new medical images registration and mosaic algorithm based on Retinex and SURF algorithms. Firstly, the algorithm took three edge filter to improved the multi scale Retinex method to enhance the images, it also used the phase correlation power spectrum to estimate the overlap region and reduced the range of registration. Then the algorithm used the SURF local feature extraction algorithm to extract and purify the feature points and established the image transform matrix to achieve mosaic. At last, according to the phenomenon of joint, this paper took lifting wavelet transform multi resolution fusion algorithm to make the image joint seemed smooth and natural. Experiments and the quality evaluation standard show that the new algorithm has high robustness to complex environment.