Efficient and accurate image registration is crucial for advancing medical image processing. However, inherent limitations in medical imaging, such as low image quality, missing data, and inconsistent visual features, are widespread and significantly affect registration accuracy. To address the challenges of registering unlabeled datasets with large deformation fields and inconsistent imaging regions, we propose a novel multimodal image registration framework that integrates both spatial and frequency domain approaches. We start by enhancing edge features in image pairs using a logarithm-weighted phase congruency edge detection method, followed by coarse registration using phase correlation methods. Subsequently, pixel-level non-local blind image denoising is applied to remove noise from registered images, improving alignment precision. We then employ a multi-feature optimized spatial registration approach to refine alignment further. Finally, applying the transformation parameters to the results of coarse registration achieves precise alignment of multimodal medical images. Our method has been successfully applied to diverse medical image types including CT and CBCT, as well as CT and MRI, demonstrating excellent registration performance and validating the effectiveness and broad applicability of the proposed model.