Chest X-ray images enhancement based on adaptive dual fractional-order differentiation kernel

Aiming at the issues of low contrast and limited dynamic range in chest X-ray images, this study proposes an adaptive dual fractional-order differentiation enhancement algorithm. We develop pixel-level adaptive order functions based on the GL-Riesz dual fractional-order kernel, which is derived from the Grünwald-Letnikov (G-L) and Riesz fractional-order kernels. Considering the necessity to enhance high-frequency components, such as rib edges, while maintaining low-contrast textural characteristics in alveolar regions, the dual-fractional kernel adaptive functions integrate multidimensional features, including gradient magnitude, local variance, local entropy, and global entropy. The functions dynamically assign differentiation orders for dual-kernel across distinct regions, achieving edge enhancement in pulmonary structures while preserving detail in smoother areas. This approach facilitates improved discernibility of lung textures and subtle pathological manifestations. The constructed adaptive dual fractional-order differential operator was applied to the COVID-19 Radiography Database dataset for evaluation. The experimental results indicate that the algorithm enhances image contrast and details. Both qualitative and quantitative analyses confirm the performance advantages of the algorithm.