Phase images of magnetic resonance imaging (MRI) have applications in many fields, including the medical domain. It is often employed to identify biomarkers of neurodegenerative diseases such as Alzheimer, Parkinson, and others. However, directly extracted phase images from MRI exhibit the wrapped phase values within the \(\pm \pi \) radian range. To circumvent these phase jumps or discontinuities, phase unwrapping is required. Path-following and minimum-norm algorithms are unwrapping methods for retrieving the original unwrapped phase image. The path-following algorithm extracts the original phase value by considering the adjacent pixels along the integral path. In contrast, the minimum-norms algorithm aims to minimize the difference between the partial derivatives of the wrapped and unwrapped phase data. This paper presents a discrete cosine transform (DCT)-based modified minimum norm-based weighted least square (WLS) phase unwrapping to improve the visibility and noise immunity of phase images. The proposed algorithm suppresses high-frequency residual noise by imposing spectral truncation of the high-frequency coefficient. For the experimental validation of the proposed method, performance is compared in terms of peak signal-to-noise ratio (PSNR) and structural similarity index metric (SSIM). The proposed phase unwrapping method outperforms state-of-the-art techniques.
