Multi-contrast magnetic resonance imaging (MRI) super-resolution (SR) typically leverages easily accessible high-resolution (HR) MRI sequence images to guide the rapid reconstruction of HR images from low-resolution (LR) MRI sequence images. However, traditional methods extract high-frequency texture information from HR images and directly integrate it into LR images for SR. These methods struggle to effectively capture subtle anatomical features when dealing with complex anatomical structures and tend to introduce redundant texture information from HR images, leading to the generation of artifacts in the super-resolved images. To address this issue, we propose a multi-contrast MRI SR method based on spatial-frequency domain cascade fusion (SFCFNet). Specifically, we first introduce a Gabor transform block with variable window-width to focus on local joint spatial-frequency features, enabling precise extraction of fine texture structures within complex anatomical regions. Then, we design cascade feature fusion module that adaptively merges complementary high-frequency detail features, suppressing redundant information and generating high-fidelity super-resolved images. Extensive experiments on the IXI and BraTS2020 datasets demonstrate that our method outperforms several state-of-the-art methods, validating its effectiveness.