This letter proposes a novel image edge detection algorithm based on the difference of two synchronizing spatially extended nonlinear dynamical systems. The time evolution of each system is identified as a low-pass filtering process, whereas the difference of two synchronizing states is identified as the result of high-pass filtering. Two systems are weakly coupled and allowed to take a common image as their initial value, but have different time scales in their dynamical motions. Results are compared with those of standard image processing schemes. Some of the best image edges could be obtained by an appropriate balance between synchronization and desynchronization.