The implementation of a visual motion detection algorithm on an analog network is presented. The algorithm is based on Markov random field (MRF) modeling. Robust motion detection is achieved by using a spatiotemporal neighborhood for modeling pixel interactions. Not only are the moving edges detected, but also the inner part of moving regions. Moreover, the motion echo is eliminated. For hardware implementation, the algorithm is mapped onto a resistive network incorporating, at each pixel-node, light acquisition and on-chip processing. Elementary Markov cells are made of resistors and non-linear current sources driven by observations, derived from luminance information acquired by photoreceptors. Electrical simulations of the cell are reported. Experimental results with synthetic and real-world image sequences exhibit the performance of the network. For circuit design, the switched-current technique is used. The circuit is intended for submicronic complementary metal-oxide-semiconductor (CMOS) technology.