Methods and apparatus for digital data array edge enhancement are disclosed. A local data window containing a data sample s is selected. Minimum and maximum sample values, max and min, are located within the window, and an edge deflection value ed is defined to have a value between max and min. A diffusion quantity is then calculated to move the value of s towards max, if s is greater than ed, or towards min, if s is smaller than ed. This approach has advantages over gradient-based edge-enhancement, including simplicity, convergence speed, and stability.