Method and apparatus for edge-based spatio-temporal filtering

An apparatus for forming super-resolution pictures from a sequence of pictures, comprising:

• a spatial filter for spatially filtering a picture in the sequence and at least one reference picture selected from among previous pictures and subsequent pictures in the sequence with respect to the picture;
• a motion compensator, in signal communication with said spatial filter, for forming, subsequent to spatial filtering, multiple temporal predictions for the picture from the at least one reference picture;
• a spatial upsampler, in signal communication with said motion compensator, for performing spatial upsampling on the picture and the at least one reference picture;
• a deblocking filter, in signal communication with said motion compensator, for deblock filtering the multiple temporal predictions;
• and
• a temporal filter, in signal communication with said deblocking filter, for temporally filtering the multiple temporal predictions and combining the multiple temporal predictions to generate an upsampled version of the picture.

A method for filtering a sequence of pictures, comprising:

• spatially filtering a picture in the sequence and at least one reference picture selected from among previous pictures and subsequent pictures in the sequence with respect to the picture;
• forming , subsequent to spatial filtering, multiple temporal predictions for the picture from the at least one reference picture;
• deblock filtering the multiple temporal predictions;
• and
• temporally filtering the multiple temporal predictions and combining the multiple temporal predictions to generate a noise reduced version of the picture, subsequent to said deblock filtering.

A method for forming super-resolution pictures from a sequence of pictures, comprising:

• spatially filtering a picture in the sequence and at least one reference picture selected from among previous pictures and subsequent pictures in the sequence with respect to the picture;
• forming subsequent to spatial filtering, multiple temporal predictions for the picture from the at least one reference picture;
• performing spatial upsampling on the picture and the at least one reference picture;
• deblock filtering the multiple temporal predictions;
• and
• temporally filtering the multiple temporal predictions and combining the multiple temporal predictions to generate an upsampled version of the picture, subsequent to said deblock filtering.