Provided is a method for prediction coding using scalable video codec, which can support various chrominance spaces. A method for prediction coding includes setting a target layer based on corresponding device information on a specific image service request, performing a chrominance space conversion on an image signal of a base layer using a chrominance space conversion equation preset in a scalable video codec selectively according to whether a chrominance space conversion is necessary in prediction coding between the base layer and the target layer, and performing inter-layer prediction coding through an additional signal conversion based on necessary any one of spatial scalability, chroma-format scalability and bit-depth scalability.