Yiting Shao authored Dec 07, 2018 and David Flynn committed Feb 06, 2019

This partitioning method (--partitionMethod=2) finds the longest edge of
the point cloud and divides it into --partitionNumUniformGeom=n slices
along the longest edge.

If n = 0, the ratio of longest edge to shortest edge determines
the number of slices.

This commit provides a basic frame work for partitioning a frame into
multiple slices, with the continued assumption of single-frame
sequences.

The decoder is modified to independently decode each slice and
accumulate decoded points in a buffer for output.

The encoder is updated to support partitioning the input point cloud
into slices and to independently code each slice.  Points from
reconstructed slices are accumulated and output at the end of the
frame period.

The partitioning process (partitioning methods are defined in
partitioning.cpp) proceeds as follows:

 - quantise the input point cloud without duplicate point removal
   or reordering points.

 - apply the partitioning function to produce a list of tiles
   and slices, each slice having an origin, id, and list of point
   indexes that identify points in the input point cloud.

 - producing a source point cloud for each partition as a subset
   of the input point cloud.

 - compressing each partition (slice) as normal by quantising
   the partitioned input.  Recolouring is necessarily performed
   against the partitioned input since the recolouring method
   cannot correctly handle recolouring a partition from a complete
   point cloud.

NB: this commit does not provide any partitioning methods.

Vladyslav Zakharchenko authored Dec 14, 2018 and David Flynn committed Feb 06, 2019

This uses a binary-tree to generate the levels-of-details for the
lifting/predicting transforms as an alternative to Euclidean distance
thresholding methods.

Jungsun Kim authored Jan 30, 2019 and David Flynn committed Feb 05, 2019

This simplified attribute prediction scheme avoids computing the
LoD structure in order to reduce the computational complexity of
both the encoder and the decoder.  It is activated by setting
levelOfDetailCount=0.

NB: this is only configured for lossless/near-lossless predictive
attribute coding of cat3 sequences.

The configuration parameters seq_bounding_box_xyz0 and
seq_bounding_box_whd allow overriding the automatic derivation of
frame (sequence) bounding box, forcing the depth of the geometry
tree.

NB: seq_bounding_box_xyz0 and seq_bounding_box_whd are specified
    using unscaled co-ordinates.

NB: this patch does not signal these parameters in the SPS.

Define po-lite option traits to handle PCCVector3<T> as a container
to permit using PCCVector3<T> types as command / config parameters.

This commit provides trisoup_node_size_log2 as a replacement for
the following HLS syntax variables:
 - geom_codec_type (redundant wrt. triangle node size derivation)
 - trisoup_depth (redundant wrt. bounding box)
 - trisoup_triangle_level (semantics simplified)

The following config options have been removed:
 - geometryCodec (redundant wrt. trisoup_node_size_log2)
 - triSoupLevel  (replaced by trisoup_node_size_log2)
 - triSoupDepth  (unnecessary)
 - triSoupIntToOrigScale (redundant wrt. positionQuantizationScale)

Khaled Mammou authored Sep 22, 2018 and David Flynn committed Nov 01, 2018

Replace the k-d tree based nearest neighbour search with an approximate
search based upon the points in Morton order.

Toshiyasu Sugio authored Sep 20, 2018 and David Flynn committed Nov 01, 2018

This commit provides an ability to vary the number of direct neighbours
used in the predicting transform.  Mode 0 corresponds to the existing
averaging, and mode n selects the n-th nearest neighbour for direct
prediction.

The provided code has been reworked to:
 - rewrite unary coding to be a little more obvious
 - tidy the mode decision path
 - add configuration for max_num_direct_predictors (renamed from
   MaxNumPredCand)
 - now counts the number of single predictors, rather than prediction
   modes.

This commit provides a method to predict the child occupancy bits of a
node based on the node's 26 neighbours.  The prediction is used to
contextualise coding of each occupancy bit.

This tool requires the use of the occupancyAtlas for neighbour lookup.
NB: a restriction in the current implementation requires that the
atlas size is at most 8³.

intra_pred_max_node_size_log2: 6

Bitwise geometry occupancy coding uses a mapping table to select entropy
contexts.  This commit employs a dynamic mapping which is updated after
each coding operation, replacing the previous static mapping tables.

NB: the proposed version used a context with a halving period
(max_count) of 64 symbols.  However, this conflicts with another
adoption (512 symbols), and a wholesale replacement of the arithmetic
codec and context model.  To resolve the conflict, the existing halving
period (128) is used.

Khaled Mammou authored Sep 19, 2018 and David Flynn committed Nov 01, 2018

This commit allows bytewise encoding of occupancy symbols using the
dual-lut coder.

Ohji Nakagami authored Jul 26, 2018 and David Flynn committed Oct 31, 2018

When trisoup was completely independent, trisoup handled lossless
geometry coding by using a leaf size of 1 and no triangles.  This
commit replaces that mode with the better performing octree geometry
coder.

This commit removes unnecessary include directives.

This commit moves various constants from PCCTMC3Common.h to a new
constants.h.  Hard coded values of constants have been replaced with
their symbolic name.

Sometimes it is useful to encode geometry without attributes.  However,
the common test condition configurations all enable attribute coding.
In order to avoid rewriting the configuration files, the
"disableAttributeCoding" option configures the encoder as if no
attributes are specified.

Reading and writing binary ply files is significantly faster than going
via formatted I/O.  This commit adds the --outputBinaryPly=1|0 option
to select writing either binary or ascii ply files.

Khaled Mammou authored Aug 27, 2018 and David Flynn committed Sep 03, 2018

Certain test sequences with 16-bit reflectance are actually 8-bit data that
has been scaled by 255.  This commit adds non-normative support to the
encoder and decoder to convert between the two representations through
the option "hack.reflectanceScale".

The dist2 values used by the pred/lift attribute coding schemes is
dependent upon the distances between points in the point cloud.  Using
the --positionQuantizationScale option to scale the input point cloud
therefore affects the required dist2 values.

This commit adds the --positionQuantizationScaleAdjustsDist2=1|0 option
to simplify the configuration of dist2.  When enabled, the encoder will
automatically scale any dist2 values by the squared position scale
factor.

Khaled Mammou authored Aug 22, 2018 and David Flynn committed Aug 29, 2018

This commit introduces automatic derivation of dist2 values by the
frontend.  Automatic derivation occurs if fewer than levelOfDetailCount
values are provided.  No syntax changes are made.

cfg update:
 - use single dist2 values based on existing config
 - ford: 16383 -> 16384

Khaled Mammou authored Aug 22, 2018 and David Flynn committed Aug 29, 2018

This commit removes the ability to vary quant_step_size on a per-LoD
basis.  This greatly simplifies the codec configuration, especially
when used in conjunction with dist2 derivation.

cfg update:
 - rename quantizationSteps* -> quantizationStep*
 - update quantizationStep* to single values:
   - lossless-geom lossy-attrs:        8, 16, 32, 64, 128, 256 [8bit]
   - lossless-geom nearlossless-attrs: no change
   - lossless-geom lossless-attrs:     no change
   - lossy-geom    lossy-attrs:        8                       [8bit]

 => 16-bit cat3 sequences are have 8-bit qs values multipled by 255

Khaled Mammou authored Aug 22, 2018 and David Flynn committed Aug 29, 2018

This commit updates the hard coded limit and default values, and
updates the master configuration template.

Khaled Mammou authored Aug 22, 2018 and David Flynn committed Aug 29, 2018

This commit replaces the hard coded threshold values with APS signalling.
The default configuration maintains the hard-coded threshold values based
on bitdepth.

This option was added to use the original CfP response output position
rounding after the switch to integer point clouds.  It is not used and
its effect is barely noticeable in the objective distortion metric.
Furthermore, the encoder and decoder behaviour is different if the
option is enabled.

This commit allows the attribute codec to be parametrised by attribute
bitdepth.  The attribute bitdepth is signalled in the attribute's sequence
parameter set description.  The specified bitdepth affects cliping in the
attribute prediction and reconstruction process.

This commit allows individual payloads to be inserted into the decoder
rather than the decoder expecting to decode each payload in a fixed
order.

When an output pointcloud is ready, the decoder will call a user
supplied callback with the conforming output pointcloud.  The user
may choose to rescale this point cloud as necessary.

This commit removes the "lossless" geometry bypass coding mode that simply
skipped coding of geometry and required the decoder to read the geometry
from the source video.  Bypass coding mode is replaced by octree coding.

This commit rewrites the codec high-level syntax:

 - the bitstream is divided into "bricks" (akin to an AVC/HEVC
   slice/tile).

 - sequence, geometry and attribute parameter sets describe the coding
   parameters in use generally and for a specific brick.

 - marshalling the bitstream payloads to a file format is achieved using
   a type-length-value encoding scheme.

Additionally, the triSoup bitstream scale and translation values have
been unified with (replaced by) the octree counterparts.  For
compatibility, existing command line parameters continue to function
as before.

NB: this commit does not incorporate flexibility in the decoding order.
The decoder requires the bitstream to be presented in a fixed order.

The IntegerLift name is misleading in light of a separate coding method
that uses lifting predict and update operators.  This commit renames
the method to specify it is a prediction-only scheme.

Khaled Mammou authored Jun 30, 2018 and David Flynn committed Aug 20, 2018

This commit implements a lifting scheme optimized for lossy attribute
compression, consisting of predict/update operators, and an adaptive
quantization strategy.

Khaled Mammou authored Jun 30, 2018 and David Flynn committed Aug 20, 2018

Adaptively pick the number of neighbours considered during the
prediction based on analysis of the reconstructed attribute values of
the neighbours.

From m42642, this is performed by computing the maximum difference
between any two reconstructed attribute values of a points neighbours.
If the variability is higher than a threshold, then apply a
rate-distortion optimization procedure to choose between using either
a single neighbour or all neighbours. The index of the best prediction
strategy is arithmetically encoded.

This commit:
 - implements the m42642 method
 - moves buildPredictors into transform-specific attribute encoding/decoding
 - adds quantizationStepsChroma
 - renames quantizationSteps -> quantizationStepsLuma

df: fixed incorrect application of quantizationStepsLuma|Chroma
df: updated cat3 cfg to specify quantizationStepsChroma
df: fixed header encoding if quantizationStepsChroma missing

This commit removes the searchRange option that was set to 0 by m42538.

In the contextualisation of occupancy coding using neighbour patterns,
this commit replaces the case when neighbour contextualisation is
disabled with a constraint that restricts neighbour contextualisation
only examine the direct siblings of a node.

The restriction ensures that at most three neighbours may be present
for a node.

The configuration option "neighbourContextualisation=1" is replaced with
"neighbourContextRestriction=0".

This commit aims to reduce te amount of copy+paste code used to
support the three different geometry coding modes by unifying them
into a single flow.

This commit adds a DecoderParameters struct to reduce the overhead of
adding extra parameters.