Abstract  Color information plays an important role in better  understanding of natural scenes by at least facilitating  discriminating boundaries of objects or areas. In this study,  we propose a new framework for boundary detection in  complex natural scenes based on the color-opponent  mechanisms of the visual system. The red-green and  blue-yellow color opponent channels in the human visual  system are regarded as the building blocks for various  color perception tasks such as boundary detection. The  proposed framework is a feedforward hierarchical model,  which has direct counterpart to the color-opponent  mechanisms involved in from the retina to the primary  visual cortex (V1). Results show that our simple framework  has excellent ability to flexibly capture both the structured  chromatic and achromatic boundaries in complex scenes.