Predictive coding using spatial-temporal filtering and plural motion vectors

A video coding apparatus, comprising:

• memory means for storing at least one of plural field picture signals and plural frame picture signals used for prediction as plural reference picture signals;
• prediction signal producing means including first filtering means for spatial filtering each of said plural reference picture signals designated by each of one set of motion vectors obtained as one set of candidates, to produce first prediction signal candidates, each of said one set of motion vectors designating each of said plural reference picture signals, and second filtering means for spatial-temporal filtering the plural reference picture signals, each of which is designated by each of said corresponding motion vectors, to produce a second prediction signal candidate;
• motion vector detecting means for detecting an optimum set of motion vectors corresponding to said one set of motion vectors, which produce a prediction signal candidate with a smallest prediction error as a prediction signal, from among the prediction signal candidates obtained from at least one of said first filtering means and said second filtering means, within plural possible sets of motion vector candidates;
• selecting means provided in said prediction signal producing means, for selecting and outputting the prediction signal candidate determined by said motion vector detecting means as the prediction signal;
• and
• coding means for coding the to-be-coded picture signal using the prediction signal.