Content adaptive bi-directional or functionally predictive multi-pass pictures for high efficiency next generation video coding

A computer-implemented method for video coding, comprising:

• receiving frames in an input video order;
• coding a first segment of a current frame in a first pass, wherein the first segment is associated with content shown on the frame, and wherein the current frame is a bi-directional (B-picture) or functional (F-picture) multi-pass picture wherein both the bi-directional and functional multi-pass picture are provided with the option to use at least one past reference frame, at least one future reference frame, or both, and wherein current, past, and future are relative to the input video order, and wherein the functional multi-pass picture has the option to use one or more modified reference frames that are modified by a morphing technique or a synthesis technique;
• coding at least a second segment of the current frame in at least a second pass and that is different than the first segment, wherein the first pass and second pass are performed at different times so that the use of reference frames during each pass occurs at different times, and wherein coding at least both the first and second segments being cooperatively used to achieve substantially a full quality coding of the substantially entire current frame;
• and
• writing data associated with reconstructing the segments to an encoder bitstream either at the same time instance or at different time instances.

A coder, comprising:

• an image buffer;
• a graphics processing unit communicatively coupled to the image buffer and comprising:
• a morphing analyzer or a synthesizing analyzer or both,
• a segmentor to divide a pixel frame into at least first and second segments associated with content in a current frame, and
• a unit to indicate which frames are to be reference frames for a segment, the graphics processing unit being configured to:
• receive frames in an input video order;
• code a first segment of a current frame in a first pass, wherein the first segment is associated with content shown on the frame, and wherein the current frame is a bi-directional (B-picture) or functional (F-picture) multi-pass picture wherein both the bi-directional and functional multi-pass picture are provided with the option to use at least one past reference frame, at least one future reference frame, or both, and wherein current, past, and future are relative to the input video order, and wherein the functional multi-pass picture has the option to use one or more modified reference frames that are modified by a morphing technique performed by the morphing analyzer or a synthesis technique performed by synthesizing analyzer;
• code at least a second segment of the current frame in at least a second pass and that is different than the first segment, wherein the first pass and second pass are performed at different times so that the use of reference frames during each pass occurs at different times, and wherein coding at least both the first and second segments being cooperatively used to achieve substantially a full quality coding of the substantially entire current frame;
• and
• write data associated with reconstructing the segments to an encoder bitstream either at the same time instance or at different time instances.