基于超小波变换的医学图像融合方法研究

发布时间：2018-01-23 03:13

本文关键词： 多模态图像融合 NSCT变换 PCNN 拉普拉斯能量和稀疏表示　出处：《江西财经大学》2016年硕士论文　论文类型：学位论文

【摘要】：现代科学技术的发展带动了医疗器械的蓬勃发展,其中的医疗影像诊断设备也络绎不绝的出现,使得医疗影像越来越多。医疗诊断中,病人需要拍摄不同医疗影像,这需要熟练专业的医生才能根据这些医疗影像诊断病人的情况,而且医生在诊断时需要综合这些医疗影像中的信息,这对医生特别是实习医生是一个很大的挑战。由此出现了医学图像融合技术,其原理是将不同的医疗影像综合在一起,把拍摄病人的医疗影像信息在一张医疗影像上显示。本文主要对医疗影像融合算法中存在的问题进行分析,通过对基于小波变换、超小波变换、邻域熵脉冲耦合神经网络(PCNN)与稀疏表示等医疗影像融合算法进行了解与思考,提出了两种新的医疗影像融合方案。工作主要创新性工作如下:1.为了提高医学图像融合的质量,提出一种基于非下采样Contourlet变换(NSCT)与邻域熵脉冲耦合神经网络(PCNN)的融合方法。在充分考虑图像NSCT域系数特征和PCNN符合人眼的视觉特性基础上,针对低、高频分别提出一种基于改进拉普拉斯能量和(Sum-Modified-Laplacian,SML)的低频系数融合方法和一种基于邻域熵的PCNN高频系数融合方法,构建一种全新的融合规则运用于医学图像融合模型中。新的融合算法能取得更好的视觉效果以及较好的客观指标值,验证了算法的优越性。2.提出了基于YUV变换和稀疏表示的结构与功能医学图像融合算法。利用YUV变换保留功能图像的颜色信息以及Y分量能够代表源图像的灰度信息的特性,结合NSCT变换实现Y分量与结构图像的融合,同时低频采用稀疏表示的方法进行低频子带融合,高频采用局部能量的方法进行融合。算法同时考虑到了色彩特征的保留以及空间分辨率的提升两个方面,取得了较好的融合结果。
[Abstract]:The development of modern science and technology has led to the vigorous development of medical devices, in which medical imaging diagnostic equipment is also in constant stream, making more and more medical images. Patients need to take different medical images, which require a skilled professional doctor to diagnose the patient according to these medical images, and the doctor needs to synthesize the information in these medical images in the diagnosis. This is a great challenge for doctors, especially interns. From this, medical image fusion technology emerged, the principle of which is to integrate different medical images together. The medical image information of the patients is displayed on a medical image. This paper mainly analyzes the problems in the medical image fusion algorithm, through the wavelet transform, super-wavelet transform. Neighborhood entropy pulse coupled neural network (PCNN) and sparse representation of medical image fusion algorithm for understanding and thinking. Two new medical image fusion schemes are proposed. The main innovative work is as follows: 1. In order to improve the quality of medical image fusion. A neural network based on non-downsampling Contourlet transform and neighborhood entropy pulse coupling neural network (PCNN) is proposed. Based on the full consideration of the image NSCT domain coefficients and the visual characteristics of the PCNN in accordance with the human eye. For low and high frequency, a new method based on improved Laplace energy and Sum-Modified-Laplacian is proposed. SML-based low frequency coefficient fusion method and a neighborhood entropy based PCNN high frequency coefficient fusion method. A new fusion rule is constructed and applied to the medical image fusion model. The new fusion algorithm can achieve better visual effect and better objective index value. The superiority of the algorithm is verified. 2. A structural and functional medical image fusion algorithm based on YUV transform and sparse representation is proposed. The color information of the functional image and the Y component can be represented by the YUV transform. The characteristics of grayscale information of an image. Combining with NSCT transform, the fusion of Y component and structure image is realized. At the same time, sparse representation is used in low frequency subband fusion. The algorithm takes into account the preservation of color features and the enhancement of spatial resolution, and achieves good fusion results.
【学位授予单位】：江西财经大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R445;TP391.41

【参考文献】