Prediction apparatus and method for improving coding efficiency in scalable video coding

Claim 2

A prediction method for improving coding efficiency by reducing temporal redundancy in an enhancement error signal of a multiple layer video encoding system, the prediction method comprising: dividing an input video sequence into a plurality of video object layers each consisting of a base layer and several enhancement layers, and each of the video object layers are further made up of a series of video object planes (VOP) comprising a two dimensional array of sampled pixel data at a particular time reference; encoding a VOP of the base layer to obtain a compressed bitstream and a corresponding locally decoded VOP of the base layer; obtaining a pixel classification criteria from information in current and previous locally decoded VOP of the base layer; constructing a merged VOP, from a previous locally decoded VOP of an enhancement layer, the previous locally decoded VOP of the base layer, the current locally decoded VOP of the base layer and the pixel classification criteria; predicting an enhancement layer VOP from: the merged VOP, the previous VOP in the same layer, and the VOP from the base layer with the same time reference; entropy coding information for prediction modes as header information for the decoder; coding prediction errors of the enhancement layer VOP which was predicted according to one of several prediction modes based on a mode decision, together with motion vector information, and transmitting them in a compressed bitstream to a decoder; and repeating the above prediction steps on the enhancement layers treating a lower of two enhancement layers as the base layer.

Claim 3

A prediction method for improving coding efficiency in multiple layer video coding according to claim 1 where obtaining the pixel classification criteria from the information in the current and previous locally decoded VOP of the base layer, further comprises: up-sampling the base layer VOP by a filtering technique, if necessary, to form a scaled VOP of the same dimension as the enhancement layer; and classifying each pixel in the current VOP as changed or unchanged based on the magnitude of the difference between the current pixel and the pixel at a corresponding location in the previous VOP.

Claim 4

A prediction method for improving coding efficiency in multiple layer video coding according to claim 2 where constructing a merged VOP, from the previous locally decoded VOP of the enhancement layer, the previous locally decoded VOP of the base layer, the current locally decoded VOP of the base layer and the pixel classification criteria, further comprises: selecting a first group of pixels for the merged VOP from the previous locally decoded VOP of the enhancement layer VOP at pixel locations classified as unchanged; selecting a second group of pixels for the merged VOP from the up-sampled VOP of the same time reference, at pixel locations classified as changed; and constructing the merged VOP by merging the groups of first and second selected pixels together.

Claim 5

A prediction method for improving coding efficiency in multiple layer video coding according to claim 3 where the prediction of the enhancement layer VOP further comprises: generating motion vectors from the previous locally decoded VOP of the enhancement layer, to be used for a forward motion compensation prediction mode; generating the motion vectors from the up-sampled VOP with the same time reference as the current predicted VOP, to be used a backward motion compensation prediction mode; searching for and obtaining a forward motion compensated macro block from previous locally decoded VOP of the enhancement layer, by using the forward motion vectors generated in the forward prediction mode for a current macro block; searching for and obtaining a backward motion compensated macro block from the up-sampled VOP, of the same time reference as the current VOP, by using the backward motion vectors generated in the backward prediction mode for the current macro block; averaging the forward motion compensated macro block and backward motion compensated macro block to obtain an interpolated macro block, for an interpolated prediction mode; and obtaining a macro block from the merged VOP for the merge prediction mode.

Claim 6

A prediction method according to claim 4 where the prediction of the enhancement layer further comprises: calculating the absolute difference between the macro block from the current VOP and the corresponding forward motion compensated macro block of the forward prediction mode; calculating the absolute difference between the macro block from the current VOP and the corresponding backward motion compensated macro block of the backward prediction mode; calculating the absolute difference between the macro block from the current VOP and the corresponding interpolated macro block of the interpolated prediction mode; calculating the absolute difference between the macro block from the current VOP and the corresponding merged macro block in the merged prediction mode; selecting the prediction mode which results in the minimum absolute difference; and predicting each of the macro blocks by using the selected prediction mode.

Claim 7

A prediction method according to claim 5 where the prediction of the enhancement layer is biased towards selecting the merged mode.

Claim 8

A prediction method for improving coding efficiency in multiple layer video coding according to claim 2 where classifying each pixel in the current VOP into changed and unchanged, further comprises: comparing the magnitude of the difference to a predefined threshold; classifying the pixel as unchanged when the magnitude of the difference is less than or equal to the threshold; and classifying the pixel as changed when the magnitude of the difference is greater than the threshold.

Claim 9

A prediction method according to claim 4 wherein merging of the groups of selected pixels employs a smoothing filter or weighting function for pixels at a boundary between the groups of selected pixels.

Claim 10

A prediction method for improving coding efficiency in multiple layer video coding according to claim 5 where the prediction mode information is entropy coded, comprising: obtaining statistics of the prediction mode selection; assigning less bits for the prediction mode with the higher possibility of being selected, and more bits for the prediction mode with the lower possibility of being selected; and assigning a default made for macro-blocks that are not transmitted as the merged prediction mode with no prediction error coded.

Claim 11

A prediction method for improving coding efficiency in multiple layer video coding according to claim 1 where obtaining the pixel classification criteria from the information in the current and previous locally decoded VOP of the base layer, further comprises: up-sampling the base layer VOP by a filtering technique, if necessary, to form a scaled VOP of the same dimension as the enhancement layer; and classifying each pixel in the current VOP as changed or unchanged based on the magnitude of the difference between the weighted sum of the current and surrounding pixels and the weighted sum of the pixels at corresponding locations in the previous VOP.

Claim 12

A prediction method for improving coding efficiency in multiple layer video coding according to claim 10 where constructing a merged VOP, from the previous locally decoded VOP of the enhancement layer, the previous locally decoded COP of the base layer, the current locally decoded VOP of the base layer and the pixel classification criteria, further comprises: selecting a first group of pixels for the merged VOP from the previous locally decoded VOP of the enhancement layer VOP at pixel locations classified as unchanged; selecting a second group of pixels for the merged VOP from the up-sampled VOP of the same time reference at pixel locations classified as changed; and constructing the merged VOP by merging the groups of first and second selected pixels together.

Claim 13

A prediction method for improving coding efficiency in multiple layer video coding according to claim 11 where the prediction of the enhancement layer VOP further comprises: generating motion vectors from the previous locally decoded VOP of the enhancement layer, to be used for a forward motion compensation prediction mode; generating the motion vectors from the up-sampled VOP with the same time reference as the current predicted VOP, to be used for a backward motion compensation prediction mode; searching for and obtaining a forward motion compensated macro block from the previous locally decoded VOP of the enhancement layer, by using the forward motion vectors generated in the forward prediction mode for a current macro block; searching for and obtaining a backward motion compensated macro block from the up-sampled VOP, of the same time reference as the current VOP, by using the backward motion vectors generated in the backward prediction mode for the current macro block; averaging the forward motion compensated macro block and backward motion compensated macro block to obtain an interpolated macro block, for an interpolated prediction mode; and obtaining a macro block from the merged VOP for the merge prediction mode.

Claim 14

A prediction method according to claim 12 where the prediction of the enhancement layer further comprises: calculating the absolute difference between the macro block from the current VOP and the corresponding forward motion compensated macro block of the forward prediction mode; calculating the absolute difference between the macro block from the current VOP and the corresponding backward motion compensated macro block of the backward prediction mode; calculating the absolute difference between the macro block from the current VOP and the corresponding interpolated macro block of the interpolated prediction mode; calculating the absolute difference between the macro block from the current VOP and the corresponding merged macro block in the merged prediction mode; selecting the prediction mode which results in the minimum absolute difference; and predicting each of the macro blocks by using the selected prediction mode.

Claim 15

The prediction method according to claim 13 where the prediction of the enhancement layer is biased towards selecting the merged mode.

Claim 16

A prediction method for improving coding efficiency in multiple layer video coding according to claim 10, where classifying each pixel in the current VOP into changed and unchanged, further comprises: comparing the magnitude of the difference value to a predefined threshold; classifying the pixel as unchanged when the magnitude of the difference is less than or equal to the threshold; and classifying the pixel as changed when the magnitude of the difference is greater than the threshold.

Claim 17

A prediction method according to claim 11 where merging of the groups of selected pixels further employs a smoothing filter or weighting function for pixels at a boundary between the groups of selected pixels.

Claim 18

A prediction method for improving coding efficiency in multiple layer video encoding and decoding according to claim 1, where the constructing of the merged VOP is done at the macro block level.

Claim 19

A prediction method for improving coding efficiency in multiple layer video encoding and decoding according to claim 1, where the prediction mode consists of a merge mode only.

Claim 20

A prediction method for improving coding efficiency in multiple layer video encoding and decoding according to claim 1, where the prediction mode consists of a merge mode and at least one of a forward, backward and interpolated modes.

Claim 21

A prediction method according to claim 1 where the VOP consists of arbitrarily shaped video objects.

Claim 22

A prediction method according to claim 1 where the VOP consists of rectangular shaped video frames.

Claim 23

A prediction method for improving coding efficiency by reducing temporal redundancy in an enhancement error signal of a multiple layer video decoding system, the method comprising: decoding a VOP of a base layer from a compressed bitstream to obtain a corresponding decoded VOP of the base layer; obtaining a pixel classification criteria from information in current and previous decoded VOP of the base layer; constructing a merged VOP, from a previous decoded VOP of an enhancement layer, the previous decoded VOP of the base layer, the current decoded VOP of the base layer and the pixel classification criteria; decoding information for prediction modes from header information transmitted by the encoder; decoding prediction errors of an enhancement layer VOP which was predicted according to a transmitted prediction mode, together with motion vector information received in the compressed bitstream; reconstructing the enhancement layer VOP from the merged VOP, the previous VOP in the same layer, and the VOP from the base layer with a same time reference; and repeating the above reconstruction steps on the enhancement layer treating a lower of two enhancement layers as the base layer.

Claim 24

A prediction method for improving coding efficiency by reducing temporal redundancy in an enhancement error signal of a multiple layer video decoding system, the method comprising: decoding a VOP of a base layer from a compressed bitstream to obtain a corresponding decoded VOP of the base layer; obtaining a pixel classification criteria from information in current and previous decoded VOP of the base layer; constructing a merged VOP, from a previous decoded VOP of an enhancement layer, the previous decoded VOP of the base layer, the current decoded VOP of the base layer and the pixel classification criteria; decoding information for prediction modes from header information transmitted by the encoder; decoding prediction errors of an enhancement layer VOP which was predicted according to a transmitted prediction mode, together with motion vector information received in the compressed bitstream; reconstructing the enhancement layer VOP from the merged VOP, the previous VOP in the same layer, and the VOP from the base layer with a same time reference; and repeating the above reconstruction steps on the enhancement layer treating a lower of two enhancement layers as the base layer.

Claim 25

A prediction method for improving coding efficiency in multiple layer video decoding according to claim 22 where obtaining the pixel classification criteria from the information in the current and previous decoded VOP of the base layer, further comprises: up-sampling the base layer VOP by a filtering technique, if necessary, to form a scaled VOP of the same dimension as the enhancement layer; and classifying each pixel in the current VOP as changed or unchanged based on the magnitude of the difference between the current pixel and the pixel at a corresponding location in the previous VOP.

Claim 26

A prediction method for improving coding efficiency in multiple layer video decoding according to claim 23 where constructing a merged VOP, from the previous decoded VOP of the enhancement layer, the previous decoded VOP of the base layer, the current decoded VOP of the base layer and the pixel classification criteria, further comprises: selecting a first group of pixels for the merged VOP from the previous decoded VOP of the enhancement layer VOP at pixel locations classified as unchanged; selecting a second group of pixels for the merged VOP from the up-sampled VOP of the same time reference, at pixel locations classified as changed; and constructing the merged VOP by merging the groups of first and second selected pixels together.

Claim 27

A prediction method for improving coding efficiency in multiple layer video coding according to claim 24 where the reconstruction of the enhancement layer VOP further comprises: using the decoded information of the prediction mode and motion vectors to obtain one of the following prediction macro-blocks, a forward motion compensated macro block from the previous locally decoded VOP of the enhancement layer, a backward motion compensated macro block from the up-sampled VOP, of the same time reference as the current VOP, an interpolated macro block which is the average of the forward motion compensated macro block and backward motion compensated macro block, or a macro block from the merged VOP; and adding the obtained prediction macro-block to a decoded prediction difference to obtain a reconstructed macro-block.

Claim 28

A prediction method according to claim 13 where merging of the groups of selected pixels further employs a smoothing filter or weighting function for pixels at a boundary between the groups of selected pixels.

Claim 29

A prediction method for improving coding efficiency in multiple layer video decoding according to claim 22 where obtaining the pixel classification criteria from the information in the current and previous decoded VOP of the base layer, further comprises: up-sampling the base layer VOP by a filtering technique, if necessary, to form a scaled VOP of the same dimension as the enhancement layer; and classifying each pixel in the current VOP as changed or unchanged based on the magnitude of the difference between the weighted sum of the current and surrounding pixels and the weighted sum of the pixels at corresponding locations in the previous VOP.

Claim 30

A prediction method for improving coding efficiency in multiple layer video decoding according to claim 27 where constructing a merged VOP, from the previous decoded VOP of the enhancement layer, the previous decoded VOP of the base layer, the current decoded VOP of the base layer and the pixel classification criteria, further comprises: selecting a first group of pixels for the merged VOP from the previous decoded VOP of the enhancement layer VOP at pixel locations classified as unchanged; selecting a second group of pixels for the merged VOP from the up-sampled VOP of the same time reference at pixel locations classified as changed; and constructing the merged VOP by merging the groups of first and second selected pixels together.

Claim 31

A prediction method for improving coding efficiency in multiple layer video coding according to claim 28 where the reconstruction of the enhancement layer VOP, further comprises: using the decoded information of the prediction mode and motion vectors to obtain one of the following prediction macro-blocks, a forward motion compensated macro block from the previous locally decoded VOP of the enhancement layer, a backward motion compensated macro block from the up-sampled VOP, of the same time reference as the current VOP, an interpolated macro-block which is the average of the forward motion compensated macro block and backward motion compensated macro block, or a macro block from the merged VOP; and adding the obtained prediction macro-bock to a decoded prediction difference to obtain a reconstruction macro-bock.

Claim 32

A prediction method according to claim 28 where merging of the groups of selected pixels further employs a smoothing filter or weighting function for pixels at a boundary between the groups of selected pixels.

Claim 33

A prediction method for improving coding efficiency in multiple layer video encoding and decoding according to claim 21, where the constructing of the merged VOP is done at the macro block level.

Claim 34

A prediction method for improving coding efficiency in multiple layer video encoding and decoding according to claim 21, where the prediction mode consists of a merge mode only.

Claim 35

A prediction method for improving coding efficiency in multiple layer video encoding and decoding according to claim 21, where the prediction mode consists of the merge mode and at least one of a forward, backward and interpolated modes.

Claim 36

A prediction method according to claim 21 where the VOP consists of arbitrarily shaped video objects.

Claim 37

A prediction method according to claim 21 where the VOP consists of rectangular shaped video frames.

Claim 38

A prediction apparatus for improving coding efficiency by reducing the temporal redundancy in a enhancement error signal of a multiple layer video encoding system, the apparatus comprising: a divider which divides an input video sequence into a plurality of video object layers consisting of a base layer and several enhancement layers, and each video object layer further comprises a series of video object planes (VOP) comprising a two dimensional array of sampled pixel data at a particular time reference; an encoder which encodes a VOP of the base layer to obtain a compressed bitstream and a corresponding locally decoded VOP of the base layer, an obtaining part which obtains a pixel classification criteria from information in the current and previous locally decoded VOP of the base layer; a constructing part which constructs a merged VOP, from a previous locally decoded VOP of an enhancement layer, the previous locally decoded VOP of the base layer, the current locally decoded VOP of the base layer and the pixel classification criteria; a prediction part which predicts an enhancement layer VOP from: the merged VOP, the previous VOP in the same layer, and the VOP from the base layer with the same time reference; a coder which entropy codes information for prediction modes as header information for the decoder; and a coder which codes prediction errors of the enhancement layer VOP which was predicted according to one of several prediction modes based on a mode decision, together with motion vector information, and transmitting them in a compressed bitstream to a decoder; wherein the prediction apparatus further operates by treating a lower of two enhancement layers as the base layer.

Claim 39

A prediction apparatus for improving coding efficiency by reducing the temporal redundancy in a enhancement error signal of a multiple layer video encoding system, the apparatus comprising: a divider which divides an input video sequence into a plurality of video object layers consisting of a base layer and several enhancement layers, and each video object layer further comprises a series of video object planes (VOP) comprising a two dimensional array of sampled pixel data at a particular time reference; an encoder which encodes a VOP of the base layer to obtain a compressed bitstream and a corresponding locally decoded VOP of the base layer, an obtaining part which obtains a pixel classification criteria from information in the current and previous locally decoded VOP of the base layer; a constructing part which constructs a merged VOP, from a previous locally decoded VOP of an enhancement layer, the previous locally decoded VOP of the base layer, the current locally decoded VOP of the base layer and the pixel classification criteria; a prediction part which predicts an enhancement layer VOP from: the merged VOP, the previous VOP in the same layer, and the VOP from the base layer with the same time reference; a coder which entropy codes information for prediction modes as header information for the decoder; and a coder which codes prediction errors of the enhancement layer VOP which was predicted according to one of several prediction modes based on a mode decision, together with motion vector information, and transmitting them in a compressed bitstream to a decoder; wherein the prediction apparatus further operates by treating a lower of two enhancement layers as the base layer.