Abstract: A video/image encoding/decoding system based on spatial prediction, in which at least one spatial prediciton mode is used to predict for encoding or decoding. At least one prediction mode acts on at least one pixel in the region to be encoded or decoded to generate the predictive values, using a prediction filter based on one or more reference pixels; furthermore, at least one of the spatial prediction modes has one or a combination of the following characteristics: the prediction filter which is content adaptively calculated through maximizing or minimizing the optimization objective in a specific region is used to generate the predictive values. The information of at least one spatial prediction filter of a specific spatial prediction mode is coded and written into the bit stream. The encoding/decoding system of this disclosure can better capture the dynamics of different video and image contents with improved precision of spatial prediction, and better coding performance is achieved.