Video Encoding - Compressed motion vector field utilizing predictive motion coding (Nokia) - Issued Patent - PRIOR ART REQUEST

Question

Seeking Prior Art on Nokia patent US6735249

This simply describes using two obvious techniques (motion compression and fractal encoding) at the same time. Both had been around for over a decade.

Priority Date: Seeking Prior Art from before Aug 11, 1999

Claim 1 Improvement of a method for motion compensated prediction of the current video frame

In a method of operating on a video sequence, said video sequence being formed of at least a current video frame and a reference video frame, the current video frame comprising at least one first neighboring segment and a second neighboring segment, an improvement of a method for motion compensated prediction of the current video frame comprising the steps of:

Improving compression between two "video frames" by looking at how pixels move.

1. Retrieving a previously stored first motion field model, said first motion field model being a model of a first motion vector field describing the displacements of pixels in the first neighboring segment with respect to pixels in the reference video frame;

Look at how the pixels moved in a specific location in the frame.

1. determining a second motion vector field describing displacements of pixels in the second neighboring segment of the current video frame with respect to pixels in the reference video frame;

Look at how the pixels moved in a neighboring location.

1. modeling said second motion vector field using a motion model to form a second motion field model;

2. approximating said second motion field model on the basis of said first motion field model to form a prediction field model;

Compare how the pixels in the first area moved to how pixels in the neighboring location moved.

1. comparing said second motion field model with said prediction field model and forming a refinement field model, said refinement field model representing the difference between said second motion field model and said prediction field model;

Figure out the difference between how pixels moved in the two locations.

1. constructing an alternative:model representation of said second motion field model by making a summation of said prediction field model and said refinement field model;

2. calculating a first cost function wherein said first cost function includes a measure of a first image distortion incurred and a measure of a first amount of data required when using said second motion field model;

3. calculating a second cost function wherein said second cost function includes a measure of a second image distortion incurred and a measure of a second amount of data required when using said alternative model representation of said second motion field;

4. comparing said first and second cost functions and determining which of said first and second cost functions has a smaller absolute value; and

5. choosing that alternate one of said second motion field model and said alternative model representation of said second motion vector field associated with said smaller absolute value to indicate a chosen motion field model and storing said chosen motion field model.

Figure out if storing the how pixels moved in both locations is smaller than storing the difference between how they moved.

Claim 25

Same compression of pixel movement information but storing the difference frames instead of between segments within a frame.

A method of encoding a video sequence comprising a plurality of video frames said method comprising the steps of:

1. forming a first motion field model for a segment of a current video frame of the video sequence using motion-compensated prediction with respect to a previously-encoded frame of the video sequence;

2. forming a prediction motion field model for the segment of the current video frame based on a motion field model determined for an adjacent previously-encoded segment of the current frame;

3. determining a difference between the first motion field model for the segment and the prediction motion field model for the segment to obtain refinement information for the segment;

4. defining a second motion field model for the segment of the current frame as a combination of the prediction motion field model and the refinement information;

5. calculating a cost function for the first and respectively the second motion field model, the cost function for a given one of the first and second motion field models taking into account both the amount of information required to represent the motion field model and the amount of image distortion introduced by using the motion field model to encode said segment of the current frame;

6. choosing either the first motion field model or the second motion field model as the motion field model for the segment based at least in part on the calculated cost functions for the first and respectively the second motion field model; and

7. encoding the segment of the current frame using the chosen motion field model to form encoded information for the segment.

Claim 45

A video encoder that uses the two compression schemes above.

A video encoder for encoding a video sequence comprising a plurality of video frames, said video encoder comprising:

1. means for forming a first motion field model for a segment of a current video frame of the video sequence using motion-compensated prediction with respect to a previously-encoded frame of the video sequence;

2. means for forming a prediction motion field model for the segment of the current video frame based on a motion field model determined for an adjacent previously-encoded segment of the current frame;

3. means for determining a difference between the first motion field model for the segment and the prediction motion field model for the segment to obtain refinement information for the segment;

4. means for defining a second motion field model for the segment of the current frame as a combination of the prediction motion field model and the refinement information; means for calculating a cost function for the first and respectively the second motion field model, the cost function for a given one of the first and second motion field models taking into account both the amount of information required to represent the motion field model and the amount of image distortion introduced by using the motion field model to encode said segment of the current frame;

5. means for choosing either the first motion field model or the second motion field model as the motion field model for the segment based at least in part on the calculated cost functions for the first and respectively the second motion field model; and means for encoding the segment of the current frame using the chosen motion field model to form encoded information for the segment.

Answer 1

This sure looks a lot like MPEG-1 video compression which was a publicly available standard in 1991. Here's the spec.

MPEG-1 included some of the following (text copied from Wikipedia's MPEG-1 Article):

P-frames exist to improve compression by exploiting the temporal (over time) redundancy in a video. P-frames store only the difference in image from the frame (either an I-frame or P-frame) immediately preceding it (this reference frame is also called the anchor frame).

So it looks like Claim 1 just segments a P-frame and calculates a diff. Claim 25 extrapolates the segmentation to two P-frames.

Fractal encoding predates this "video segmentation" patent by over a decade.

45 appear to be little more than weak extrapolations of the other two regarding selecting results and creating them. For this I believe looking at any video editing software that offered multiple "Save As..." or "Export As..." formatting options will suffice. Adobe Premiere has done this since 1991...