Method and apparatus for selectively encoding and decoding a digital motion video signal at multiple resolution levels

Claim 3

The method in accordance with claim 1 wherein step c comprises encoding at least one lower resolution level but not encoding the full resolution level of each of said selected frames.

Claim 4

The method in accordance with claim 1 in which step c comprises, for each level of resolution of said selected frame which is encoded, the steps of: i--forming a correction image comprising an array of pixel values, by subtracting pixel values in a reference image from corresponding pixel values in that level of resolution of said selected frame; ii--quantizing the pixel values in said correction image to form a quantized correction image; and iii--encoding said quantized correction image.

Claim 5

The method in accordance with claim 3 in which step i for the lowest level of resolution of said selected frame includes the step of providing a reference frame comprising an array of pixels all of which have the same value.

Claim 6

The method in accordance with claim 4 in which step i for each level of resolution higher than the lowest level of resolution of said selected frame includes the steps of: (1) forming a result image having a lower level of resolution by adding the pixel values of the quantized correction image at that lower level of resolution to the pixel values of the reference image at that lower level of resolution; and (2) forming the reference image at the higher level of resolution by expanding the result image formed in step (1) to said higher level of resolution.

Claim 7

The method in accordance with claim 3 wherein step iii comprises the steps of dividing said quantized correction image into regions and generating region parameters descriptive of the position, size and fill data for each region.

Claim 8

The method in accordance with claim 6 wherein the step of dividing said quantized correction image into regions comprises dividing said image into null and non-null regions or into null and linear fill regions.

Claim 9

The method in accordance with claim 7 wherein said region descriptions are encoded using binary tree decomposition.

Claim 10

The method in accordance with claim 8 in which said non-null regions are encoded using vector quantization or discrete cosine transform techniques.

Claim 11

The method in accordance with claim 9 wherein encoding using vector quantization comprises encoding using DYADS, QUADS, or DPCM.

Claim 12

The method in accordance with claim 1 in which step d comprises the steps of: i--dividing each non-selected frame into regions and generating region parameters descriptive of the location and size of respective regions within said frame; ii--for non-selected still frames or a first frame of a sequence of non-selected frames: (1) determining absolute fill data describing respective regions; (2) determining DPCM fill data describing respective regions for which absolute fill data cannot be determined; (3) variable length encoding said absolute fill data, said DPCM fill data and corresponding region descriptive parameters; (4) determining an estimate of decompression time from said variable length encoded absolute, DPCM fill data and corresponding region descriptive parameters; (5) providing said variable length encoded fill data and corresponding region parameters if said estimate is less than a predetermined value, and if not, repeating steps (i) and (1)-(5) using a different tolerance in determining said absolute fill data; and iii--for a succession of non-selected frames: (1) comparing pixel values in each region with corresponding pixel values in a like sized region from a previously occurring frame to develop respective pixel differences; (2) determining relative fill data descriptive of pixel differences for respective regions; (3) determining DYAD fill data for respective regions in which relative fill data cannot be determined; (4) determining DPCM fill data for respective regions in which relative fill data cannot be determined and for which DYAD fill data cannot be determined; and (5) variable length coding said relative fill data, said DYAD fill data, said DPCM fill data and corresponding region parameters.

Claim 13

The method in accordance with claim 1 wherein step c comprises the steps of: i--dividing each resolution level which is en coded into null and non-null regions; ii--encoding said regions using binary tree decomposition; iii--determining vector values describing the pixel values in said non-null regions; and iv--quantizing said vector values.

Claim 14

The method in accordance with claim 12 wherein said vector values include DYADS, QUADS and DPCM.

Claim 15

The method in accordance with claim 1 wherein step a comprises the steps of: i--forming a motion compensated difference image for each frame of said digital motion video signal; ii--determining the value of a selected parameter associated with said motion compensated difference image; iii--selecting a threshold value of said selected parameter; and iv--selecting those frames from the sequence of frames having parameter values which exceed said predetermined thresholds.

Claim 18

The method in accordance with claim 15 in which step b(2) comprises, for each level of resolution of said selected frame which is encoded, the steps of: i--forming a correction image comprising an array of pixels, by subtracting pixel values in a reference image from corresponding pixel values in that level of resolution of said selected frame; ii--quantizing the pixel values in said correction image to form a quantized correction image; and iii--encoding said quantized correction image.

Claim 19

The method in accordance with claim 16 in which step i for the lowest level of resolution of said selected frame includes the step of providing a reference frame comprising an array of pixels all of which have the same value.

Claim 20

The method in accordance with claim 17 in which step i for each level of resolution higher than the lowest level of resolution of said selected frame includes the steps of: (1) forming a result image having a lower level of resolution by adding the pixel values of the quantized correction image at that lower level of resolution to the pixel values of the reference image at that lower level of resolution; and (2) forming the reference image at the higher level of resolution by expanding the result image formed in step (1) to said higher level of resolution.

Claim 21

The method in accordance with claim 16 wherein step iii comprises the steps of dividing said quantized correction image into regions and generating region parameters descriptive of the position, size and fill data for each region.

Claim 22

The method in accordance with claim 19 wherein the step of dividing said quantized correction image into regions comprises dividing said image into null and non-null regions or into null and linear fill regions.

Claim 23

The method in accordance with claim 18 wherein said region descriptions are encoded using binary tree decomposition.

Claim 24

The method in accordance with claim 19 in which said non-null regions are encoded using vector quantization or discrete cosine transform technique.

Claim 25

The method in accordance with claim 22 wherein encoding using vector quantization comprises encoding using DYADS, QUADS, or DPCM.

Claim 26

The method in accordance with claim 15 in which step c comprises the steps of: i--dividing each non-selected frame into regions and generating region parameters descriptive of the location and size of respective regions within said frame; ii--for non-selected still frames or a first frame of a sequence of non-selected frames: (1) determining absolute fill data describing respective regions; (2) determining DPCM fill data describing respective regions for which absolute fill data cannot be determined; (3) variable length encoding said absolute fill data, said DPCM fill data and corresponding region descriptive parameters; (4) determining an estimate of decompression time from said variable length encoded absolute, DPCM fill data and corresponding region descriptive parameters; (5) providing said variable length encoded fill data and corresponding region parameters if said estimate is less than a predetermined value, and if not, repeating steps (i) and (1)-(5) using a different tolerance in determining said absolute fill data; and iii--for a succession of non-selected frames: (1) comparing pixel values in each region with corresponding pixel values in a like sized region from a previously occurring frame to develop respective pixel differences; (2) determining relative fill data descriptive of pixel differences for respective regions; (3) determining DYAD fill data for respective regions in which relative fill data cannot be determined; (4) determining DPCM fill data for respective regions in which relative fill data cannot be determined and for which DYAD fill data cannot be determined; and (5) variable length coding said relative fill data, said DYAD fill data, said DPCM fill data and corresponding region parameters.

Claim 27

The method in accordance with claim 15 wherein step b(2) comprises the steps of: i--dividing each resolution level which is encoded into null and non-null regions; ii--encoding said regions using binary tree decomposition; iii--determining vector values describing the pixel values in said non-null regions; and iv--quantizing said vector values.

Claim 28

The method of claim 25 wherein said vector values include DYADS, QUADS and DPCM.

Claim 29

The method of claim 15 wherein step a comprises the steps of: i--forming a motion compensated difference image for each frame of said digital motion video signal; ii--determining the value of a selected parameter associated with said motion compensated difference image; iii--selecting a threshold value of said selected parameter; and iv--selecting those frames from the sequence of frames having parameter values which exceed said predetermined thresholds.