Abstract
Spatially-varying intensity noise is a common source of distortion in medical images and is often associated with reduced accuracy in medical image registration. In this paper, we propose two multi-resolution image registration algorithms based on Empirical Mode Decomposition (EMD) that are robust against additive spatially-varying noise. EMD is a multi-resolution tool that decomposes a signal into several principle patterns and residual components. Our first proposed algorithm (LR-EMD) is based on the registration of EMD feature maps from both floating and reference images in various resolutions. In the second algorithm (AFR-EMD), we first extract a single average feature map based on EMD and then use a simple hierarchical multi-resolution algorithm to register the average feature maps. We then showcase the superior performance of both algorithms in the registration of brain MRIs as well as retina images. For the registration of brain MR images, using mutual information as the similarity measure, both AFR-EMD and LR-EMD achieve a lower error rate in intensity (42% and 32%, respectively) and lower error rate in transformation (52% and 41%, respectively) compared to intensity-based hierarchical registration. Our results suggest that the two proposed algorithms offer robust registration solutions in the presence of spatially-varying noise.
Highlights
We examine the ability of state-of-the-art hierarchical signal decomposition techniques in removing additive spatially varying noise in magnetic resonance (MR) images
We used the simulated MR images in the BrainWeb dataset [43,45] to evaluate the performance of our registration algorithms
Throughout this paper, we presented evidence on the advantage of using empirical mode decomposition in the registration of single modal MR images
Summary
The accurate registration and alignment of two images has been a challenging problem in a wide variety of applications such as medical image processing [1,2], remote sensing [3], biology [4–6], and computer vision [7,8]. The registration of the medical images has been widely used in tumor localization and targeting [9], organ growth studies [10]. One class is concerned with the registration based on the intensity of the images. In this approach, a similarity measure is defined to quantify the similarity of both floating (or moving) and reference (or target) images. An optimization process identifies an optimal map for the floating image to achieve the highest similarity to the reference image. In the second class of registration algorithms, a set of features such as landmarks [13], a histogram of intensity [14]
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
