Skip to content
GitLab
Commit 51f455d0 authored by Khaled Mammou's avatar Khaled Mammou Committed by David Flynn
Browse files

m42634/attr: do not vary per point the number of neighbours for prediction 

This commit uses a constant (per attribute) number of neighbours in
point attribute prediction, replacing a method where the encoder searches
for an optimal number and signals this on a per point basis.

Extracted-from: http://wg11.sc29.org/svn/repos/MPEG-I/Part5-PointCloudCompression/CE/CE0/CE0.1/trunk@1357


Reintegrated-by: David Flynn's avatar <dflynn@blackberry.com>
parent 796c7578
Hide whitespace changes
Inline Side-by-side
tmc3/PCCTMC3Common.h
View file @ 51f455d0
......@@ -109,6 +109,9 @@ struct PCCPredictor {
void computeWeights(const size_t neighborCount0) {
neighborCount = std::min(neighborCount0, maxNeighborCount);
if (!neighborCount) {
return;
}
double sum = 0.0;
for (size_t n = 0; n < neighborCount; ++n) {
neighbors[n].weight = neighbors[n].invDistance;
......
tmc3/PCCTMC3Decoder.h
View file @ 51f455d0
......@@ -225,8 +225,7 @@ class PCCTMC3Decoder3 {
}
}
int decodeReflectances(PCCBitstream &bitstream, PCCPointSet3 &pointCloud) {
const size_t pointCount = pointCloud.getPointCount();
assert(pointCount == predictors.size());
const size_t pointCount = predictors.size();
uint32_t compressedBitstreamSize = 0;
PCCReadFromBuffer<uint32_t>(bitstream.buffer, compressedBitstreamSize, bitstream.size);
......@@ -250,28 +249,20 @@ class PCCTMC3Decoder3 {
bitstream.buffer + bitstream.size);
arithmeticDecoder.start_decoder();
o3dgc::Static_Bit_Model binaryModel0;
o3dgc::Adaptive_Data_Model neighborCountModel;
neighborCountModel.set_alphabet(uint32_t(numberOfNearestNeighborsInPrediction));
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
auto &predictor = predictors[predictorIndex];
uint16_t &reflectance = pointCloud.getReflectance(predictor.index);
if (!predictor.neighborCount) {
reflectance = decodeAbsUInt32(16, arithmeticDecoder, binaryModel0);
} else {
const size_t neighborCount = arithmeticDecoder.decode(neighborCountModel) + 1;
predictor.computeWeights(neighborCount);
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = quantizationSteps[lodIndex];
const uint32_t attValue0 = decodeDiff0UInt32(
maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0, arithmeticDecoder,
multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const int64_t quantPredAttValue = predictor.predictReflectance(pointCloud);
const int64_t delta = PCCInverseQuantization(o3dgc::UIntToInt(attValue0), qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + delta;
reflectance = uint16_t(PCCClip(reconstructedQuantAttValue, int64_t(0),
int64_t(std::numeric_limits<uint16_t>::max())));
}
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = quantizationSteps[lodIndex];
const uint32_t attValue0 = decodeDiff0UInt32(
maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0, arithmeticDecoder,
multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const int64_t quantPredAttValue = predictor.predictReflectance(pointCloud);
const int64_t delta = PCCInverseQuantization(o3dgc::UIntToInt(attValue0), qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + delta;
reflectance = uint16_t(PCCClip(reconstructedQuantAttValue, int64_t(0),
int64_t(std::numeric_limits<uint16_t>::max())));
}
arithmeticDecoder.stop_decoder();
bitstream.size += compressedBitstreamSize;
......@@ -279,8 +270,7 @@ class PCCTMC3Decoder3 {
}
int decodeColors(PCCBitstream &bitstream, PCCPointSet3 &pointCloud) {
const size_t pointCount = pointCloud.getPointCount();
assert(pointCount == predictors.size());
const size_t pointCount = predictors.size();
uint32_t compressedBitstreamSize = 0;
PCCReadFromBuffer<uint32_t>(bitstream.buffer, compressedBitstreamSize, bitstream.size);
......@@ -322,53 +312,48 @@ class PCCTMC3Decoder3 {
bitstream.buffer + bitstream.size);
arithmeticDecoder.start_decoder();
o3dgc::Static_Bit_Model binaryModel0;
o3dgc::Adaptive_Data_Model neighborCountModel;
neighborCountModel.set_alphabet(uint32_t(numberOfNearestNeighborsInPrediction));
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
auto &predictor = predictors[predictorIndex];
PCCColor3B &color = pointCloud.getColor(predictor.index);
if (!predictor.neighborCount) {
for (size_t k = 0; k < 3; ++k) {
color[k] = decodeAbsUInt32(8, arithmeticDecoder, binaryModel0);
}
} else {
const size_t neighborCount = arithmeticDecoder.decode(neighborCountModel) + 1;
predictor.computeWeights(neighborCount);
const PCCColor3B predictedColor = predictor.predictColor(pointCloud);
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = quantizationSteps[lodIndex];
const uint32_t attValue0 = decodeDiff0UInt32(
maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0, arithmeticDecoder,
multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const uint32_t modelIndex = (attValue0 < PCCTMC3Diff1AdaptiveDataModelCount)
? attValue0
: PCCTMC3Diff1AdaptiveDataModelCount - 1;
const int64_t quantPredAttValue = predictedColor[0];
const int64_t delta = PCCInverseQuantization(o3dgc::UIntToInt(attValue0), qs);
const PCCColor3B predictedColor = predictor.predictColor(pointCloud);
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = quantizationSteps[lodIndex];
const uint32_t attValue0 = decodeDiff0UInt32(
maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0, arithmeticDecoder,
multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const uint32_t modelIndex = (attValue0 < PCCTMC3Diff1AdaptiveDataModelCount)
? attValue0
: PCCTMC3Diff1AdaptiveDataModelCount - 1;
const int64_t quantPredAttValue = predictedColor[0];
const int64_t delta = PCCInverseQuantization(o3dgc::UIntToInt(attValue0), qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + delta;
color[0] = uint8_t(PCCClip(reconstructedQuantAttValue, int64_t(0), int64_t(255)));
for (size_t k = 1; k < 3; ++k) {
const uint32_t attValue = decodeDiff1UInt32(
modelIndex, maxAttributeValueDiff1, adaptiveDataModelAlphabetSizeDiff1,
arithmeticDecoder, multiSymbolModelDiff1, binaryModelDiff1, binaryModel0);
const int64_t quantPredAttValue = predictedColor[k];
const int64_t delta = PCCInverseQuantization(o3dgc::UIntToInt(attValue), qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + delta;
color[0] = uint8_t(PCCClip(reconstructedQuantAttValue, int64_t(0), int64_t(255)));
for (size_t k = 1; k < 3; ++k) {
const uint32_t attValue = decodeDiff1UInt32(
modelIndex, maxAttributeValueDiff1, adaptiveDataModelAlphabetSizeDiff1,
arithmeticDecoder, multiSymbolModelDiff1, binaryModelDiff1, binaryModel0);
const int64_t quantPredAttValue = predictedColor[k];
const int64_t delta = PCCInverseQuantization(o3dgc::UIntToInt(attValue), qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + delta;
color[k] = uint8_t(PCCClip(reconstructedQuantAttValue, int64_t(0), int64_t(255)));
}
color[k] = uint8_t(PCCClip(reconstructedQuantAttValue, int64_t(0), int64_t(255)));
}
}
arithmeticDecoder.stop_decoder();
bitstream.size += compressedBitstreamSize;
return 0;
}
void buildPredictors(const PCCPointSet3 &pointCloud) {
std::vector<uint32_t> numberOfPointsPerLOD;
std::vector<uint32_t> indexes;
PCCBuildPredictors(pointCloud, numberOfNearestNeighborsInPrediction, levelOfDetailCount, dist2,
predictors, numberOfPointsPerLOD, indexes);
for (auto &predictor : predictors) {
predictor.computeWeights(numberOfNearestNeighborsInPrediction);
}
}
int decodeAttributeHeader(const std::string &attributeName, PCCBitstream &bitstream) {
......
tmc3/PCCTMC3Encoder.h
View file @ 51f455d0
......@@ -134,7 +134,6 @@ class PCCTMC3Encoder3 {
return ret;
}
buildPredictors(colorParams);
computeColorPredictionWeights(colorParams);
if (int ret = encodeColors(colorParams, bitstream)) {
return ret;
}
......@@ -150,7 +149,6 @@ class PCCTMC3Encoder3 {
return ret;
}
buildPredictors(reflectanceParams);
computeReflectancePredictionWeights(reflectanceParams);
if (int ret = encodeReflectances(reflectanceParams, bitstream)) {
return ret;
}
......@@ -192,7 +190,6 @@ class PCCTMC3Encoder3 {
return ret;
}
buildPredictors(colorParams);
computeColorPredictionWeights(colorParams);
if (int ret = encodeColors(colorParams, bitstream)) {
return ret;
}
......@@ -208,7 +205,6 @@ class PCCTMC3Encoder3 {
return ret;
}
buildPredictors(reflectanceParams);
computeReflectancePredictionWeights(reflectanceParams);
if (int ret = encodeReflectances(reflectanceParams, bitstream)) {
return ret;
}
......@@ -268,8 +264,7 @@ class PCCTMC3Encoder3 {
}
int encodeReflectances(const PCCAttributeEncodeParamaters &reflectanceParams,
PCCBitstream &bitstream) {
const size_t pointCount = pointCloud.getPointCount();
assert(pointCount == predictors.size());
const size_t pointCount = predictors.size();
uint32_t maxAttributeValueDiff0 = 0;
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
const auto &predictor = predictors[predictorIndex];
......@@ -305,32 +300,23 @@ class PCCTMC3Encoder3 {
bitstream.buffer + bitstream.size);
arithmeticEncoder.start_encoder();
o3dgc::Static_Bit_Model binaryModel0;
o3dgc::Adaptive_Data_Model neighborCountModel;
neighborCountModel.set_alphabet(
uint32_t(reflectanceParams.numberOfNearestNeighborsInPrediction));
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
const auto &predictor = predictors[predictorIndex];
if (!predictor.neighborCount) { // no prediction
encodeAbsUInt32(pointCloud.getReflectance(predictor.index), 16, arithmeticEncoder,
binaryModel0);
} else {
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = reflectanceParams.quantizationSteps[lodIndex];
arithmeticEncoder.encode(uint32_t(predictor.neighborCount - 1), neighborCountModel);
const int64_t quantAttValue = pointCloud.getReflectance(predictor.index);
const int64_t quantPredAttValue = predictor.predictReflectance(pointCloud);
const int64_t delta = PCCQuantization(quantAttValue - quantPredAttValue, qs);
const uint32_t attValue0 = uint32_t(o3dgc::IntToUInt(long(delta)));
encodeDiff0UInt32(attValue0, maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0,
arithmeticEncoder, multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const int64_t reconstructedDelta = PCCInverseQuantization(delta, qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + reconstructedDelta;
const uint16_t reconstructedReflectance = uint16_t(PCCClip(
reconstructedQuantAttValue, int64_t(0), int64_t(std::numeric_limits<uint16_t>::max())));
pointCloud.setReflectance(predictor.index, reconstructedReflectance);
}
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = reflectanceParams.quantizationSteps[lodIndex];
const int64_t quantAttValue = pointCloud.getReflectance(predictor.index);
const int64_t quantPredAttValue = predictor.predictReflectance(pointCloud);
const int64_t delta = PCCQuantization(quantAttValue - quantPredAttValue, qs);
const uint32_t attValue0 = uint32_t(o3dgc::IntToUInt(long(delta)));
encodeDiff0UInt32(attValue0, maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0,
arithmeticEncoder, multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const int64_t reconstructedDelta = PCCInverseQuantization(delta, qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + reconstructedDelta;
const uint16_t reconstructedReflectance = uint16_t(PCCClip(
reconstructedQuantAttValue, int64_t(0), int64_t(std::numeric_limits<uint16_t>::max())));
pointCloud.setReflectance(predictor.index, reconstructedReflectance);
}
uint32_t compressedBitstreamSize = arithmeticEncoder.stop_encoder();
bitstream.size += compressedBitstreamSize;
......@@ -356,8 +342,7 @@ class PCCTMC3Encoder3 {
}
int encodeColors(const PCCAttributeEncodeParamaters &colorParams, PCCBitstream &bitstream) {
const size_t pointCount = pointCloud.getPointCount();
assert(pointCount == predictors.size());
const size_t pointCount = predictors.size();
uint32_t maxAttributeValueDiff0 = 0;
uint32_t maxAttributeValueDiff1 = 0;
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
......@@ -422,137 +407,47 @@ class PCCTMC3Encoder3 {
bitstream.buffer + bitstream.size);
arithmeticEncoder.start_encoder();
o3dgc::Static_Bit_Model binaryModel0;
o3dgc::Adaptive_Data_Model neighborCountModel;
neighborCountModel.set_alphabet(uint32_t(colorParams.numberOfNearestNeighborsInPrediction));
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
const auto &predictor = predictors[predictorIndex];
const PCCColor3B color = pointCloud.getColor(predictor.index);
if (!predictor.neighborCount) { // no prediction
for (size_t k = 0; k < 3; ++k) {
encodeAbsUInt32(color[k], 8, arithmeticEncoder, binaryModel0);
}
} else {
const PCCColor3B predictedColor = predictor.predictColor(pointCloud);
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = colorParams.quantizationSteps[lodIndex];
arithmeticEncoder.encode(uint32_t(predictor.neighborCount - 1), neighborCountModel);
const int64_t quantAttValue = color[0];
const int64_t quantPredAttValue = predictedColor[0];
const PCCColor3B predictedColor = predictor.predictColor(pointCloud);
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t qs = colorParams.quantizationSteps[lodIndex];
const int64_t quantAttValue = color[0];
const int64_t quantPredAttValue = predictedColor[0];
const int64_t delta = PCCQuantization(quantAttValue - quantPredAttValue, qs);
const uint32_t attValue0 = uint32_t(o3dgc::IntToUInt(long(delta)));
const uint32_t modelIndex = (attValue0 < PCCTMC3Diff1AdaptiveDataModelCount)
? attValue0
: PCCTMC3Diff1AdaptiveDataModelCount - 1;
encodeDiff0UInt32(attValue0, maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0,
arithmeticEncoder, multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const int64_t reconstructedDelta = PCCInverseQuantization(delta, qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + reconstructedDelta;
PCCColor3B reconstructedColor;
reconstructedColor[0] =
uint8_t(PCCClip(reconstructedQuantAttValue, int64_t(0), int64_t(255)));
for (size_t k = 1; k < 3; ++k) {
const int64_t quantAttValue = color[k];
const int64_t quantPredAttValue = predictedColor[k];
const int64_t delta = PCCQuantization(quantAttValue - quantPredAttValue, qs);
const uint32_t attValue0 = uint32_t(o3dgc::IntToUInt(long(delta)));
const uint32_t modelIndex = (attValue0 < PCCTMC3Diff1AdaptiveDataModelCount)
? attValue0
: PCCTMC3Diff1AdaptiveDataModelCount - 1;
encodeDiff0UInt32(attValue0, maxAttributeValueDiff0, adaptiveDataModelAlphabetSizeDiff0,
arithmeticEncoder, multiSymbolModelDiff0, binaryModelDiff0, binaryModel0);
const int64_t reconstructedDelta = PCCInverseQuantization(delta, qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + reconstructedDelta;
PCCColor3B reconstructedColor;
reconstructedColor[0] =
const uint32_t attValue1 = uint32_t(o3dgc::IntToUInt(long(delta)));
encodeDiff1UInt32(attValue1, modelIndex, maxAttributeValueDiff1,
adaptiveDataModelAlphabetSizeDiff1, arithmeticEncoder,
multiSymbolModelDiff1, binaryModelDiff1, binaryModel0);
reconstructedColor[k] =
uint8_t(PCCClip(reconstructedQuantAttValue, int64_t(0), int64_t(255)));
for (size_t k = 1; k < 3; ++k) {
const int64_t quantAttValue = color[k];
const int64_t quantPredAttValue = predictedColor[k];
const int64_t delta = PCCQuantization(quantAttValue - quantPredAttValue, qs);
const int64_t reconstructedDelta = PCCInverseQuantization(delta, qs);
const int64_t reconstructedQuantAttValue = quantPredAttValue + reconstructedDelta;
const uint32_t attValue1 = uint32_t(o3dgc::IntToUInt(long(delta)));
encodeDiff1UInt32(attValue1, modelIndex, maxAttributeValueDiff1,
adaptiveDataModelAlphabetSizeDiff1, arithmeticEncoder,
multiSymbolModelDiff1, binaryModelDiff1, binaryModel0);
reconstructedColor[k] =
uint8_t(PCCClip(reconstructedQuantAttValue, int64_t(0), int64_t(255)));
}
pointCloud.setColor(predictor.index, reconstructedColor);
}
pointCloud.setColor(predictor.index, reconstructedColor);
}
uint32_t compressedBitstreamSize = arithmeticEncoder.stop_encoder();
bitstream.size += compressedBitstreamSize;
PCCWriteToBuffer<uint32_t>(compressedBitstreamSize, bitstream.buffer, startSize);
return 0;
}
void computeColorPredictionWeights(const PCCAttributeEncodeParamaters &attributeParams) {
const size_t pointCount = predictors.size();
double bestCost = std::numeric_limits<double>::max();
size_t bestNeighborCount = 1;
for (size_t neighborCount = 1;
neighborCount <= attributeParams.numberOfNearestNeighborsInPrediction; ++neighborCount) {
double cost = 0.0;
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
auto &predictor = predictors[predictorIndex];
if (!predictor.maxNeighborCount) {
continue;
}
predictor.computeWeights(neighborCount);
const PCCColor3B color = pointCloud.getColor(predictor.index);
const PCCColor3B predictedColor = predictor.predictColor(pointCloud);
for (int k = 0; k < 3; ++k) {
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t quantAttValue = color[k];
const int64_t quantPredAttValue = predictedColor[k];
const int64_t delta0 = PCCQuantization(
quantAttValue - quantPredAttValue,
int64_t(attributeParams.quantizationSteps[lodIndex]));
const uint32_t delta1 = uint32_t(o3dgc::IntToUInt(long(delta0)));
cost += delta1;
}
}
if (cost < bestCost) {
bestCost = cost;
bestNeighborCount = neighborCount;
}
}
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
auto &predictor = predictors[predictorIndex];
if (!predictor.maxNeighborCount) {
continue;
}
predictor.computeWeights(bestNeighborCount);
}
}
void computeReflectancePredictionWeights(const PCCAttributeEncodeParamaters &attributeParams) {
const size_t pointCount = predictors.size();
double bestCost = std::numeric_limits<double>::max();
size_t bestNeighborCount = 1;
for (size_t neighborCount = 1;
neighborCount <= attributeParams.numberOfNearestNeighborsInPrediction; ++neighborCount) {
double cost = 0.0;
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
auto &predictor = predictors[predictorIndex];
if (!predictor.maxNeighborCount) {
continue;
}
predictor.computeWeights(neighborCount);
const size_t lodIndex = predictor.levelOfDetailIndex;
const int64_t quantAttValue = pointCloud.getReflectance(predictor.index);
const int64_t quantPredAttValue = predictor.predictReflectance(pointCloud);
const int64_t delta0 = PCCQuantization(
quantAttValue - quantPredAttValue,
int64_t(attributeParams.quantizationSteps[lodIndex]));
const uint32_t delta1 = uint32_t(o3dgc::IntToUInt(long(delta0)));
cost += delta1;
}
if (cost < bestCost) {
bestCost = cost;
bestNeighborCount = neighborCount;
}
}
for (size_t predictorIndex = 0; predictorIndex < pointCount; ++predictorIndex) {
auto &predictor = predictors[predictorIndex];
if (!predictor.maxNeighborCount) {
continue;
}
predictor.computeWeights(bestNeighborCount);
}
}
void buildPredictors(const PCCAttributeEncodeParamaters &attributeParams) {
std::vector<uint32_t> numberOfPointsPerLOD;
......@@ -560,7 +455,13 @@ class PCCTMC3Encoder3 {
PCCBuildPredictors(pointCloud, attributeParams.numberOfNearestNeighborsInPrediction,
attributeParams.levelOfDetailCount, attributeParams.dist2,
predictors, numberOfPointsPerLOD, indexes);
for (auto &predictor : predictors) {
predictor.computeWeights(
attributeParams.numberOfNearestNeighborsInPrediction);
}
}
void reconstructedPointCloud(const PCCTMC3Encoder3Parameters &params,
PCCPointSet3 *reconstructedCloud) {
if (reconstructedCloud == nullptr) {
......
Supports Markdown
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment