Multi-feature adaptive Laplacian pyramid fusion for low-contrast medical image enhancement

Problem Low-contrast MRI and CT images are hard to interpret, and existing techniques often fail to improve contrast and highlight edges. This paper proposes multi-feature linear Laplacian pyramid fusion (MFL-LPF) for joint contrast and edge enhancement.Approach Two enhanced images are obtained using k-means-based local histogram equalization and fast discrete curvelet transform. Both images are decomposed into a k-level Laplacian pyramid, and fusion is guided by Local Energy, Absolute Coefficient Magnitude, Gradient Magnitude, and Spatial Frequency. Hard-decision fusion retains high-frequency details, while weighted averaging preserves low-frequency brightness consistency.Results MFL-LPF achieved the best performance on 253 brain MRI images using entropy (6.95), PSNR (36.4 dB), MI (4.68), average gradient (11.25), QABF (0.96), and SSIM (0.78) with minimal artefacts.Conclusion MFL-LPF is a training-free, GPU-independent framework for efficient medical image enhancement, supporting real-time diagnostics and future multimodal fusion and tumour segmentation applications.