Method and apparatus for adaptive quantization for symmetry mesh
A method performed by at least one processor of an encoder comprises: performing, on an input 3D mesh, a symmetry detection process to estimate a partition plane. The method further comprises partitioning the input 3D mesh into a first side and a second side based on the partition plane, where the first side is opposite to the second side, and where a first vertex on the first side is symmetric to a second vertex on the second side. The method further comprises quantizing the first vertex and the second vertex in pairs to reduce a quantization error and symmetry prediction error associated with the first vertex and the second vertex.
1. A method performed by at least one processor of an encoder, the method comprising:
performing, on an input 3D mesh, a symmetry detection process to estimate a partition plane;
partitioning the input 3D mesh into a first side and a second side based on the partition plane, wherein the first side is opposite to the second side, wherein a first vertex on the first side is symmetric to a second vertex on the second side; and
quantizing the first vertex and the second vertex in pairs to reduce a quantization error and symmetry prediction error associated with the first vertex and the second vertex.
2. The method according to claim 1 , wherein the quantization error is based on an error between the first vertex and a quantization of the first vertex.
3. The method according to claim 2 , wherein the quantization error is further based on a difference between the second vertex and a quantization of the second vertex.
4. The method according to claim 3 , wherein the quantization of the second vertex is determined based on a symmetry prediction performed on the quantized first vertex and the partition plane.
5. The method according to claim 1 , wherein the quantizing further comprises performing a search in a 3D space around the quantized first vertex to find a best vertex on the first side and a best vertex on the second side that reduces the quantization error and the symmetry prediction error.
6. The method according to claim 1 , wherein the quantizing is performed on K symmetry pairs based on the partition plane, and
wherein the partition plane is updated based on the quantizing performed on the K symmetry pairs.
7. The method according to claim 6 , wherein the quantizing performed on the K symmetry pairs and the updating of the partition plane is iteratively performed until a change in the partition plane between a first iteration and a second iteration is less than a threshold.
8. The method according to claim 1 , further comprising:
determining, based on an a 2D matching process performed on the input 3D mesh that is partitioned, a 2D mesh;
segmenting the 2D mesh into a plurality of non-connected segments;
performing, for at least one non-connected segment in the plurality of non-connected segments, a symmetry detection process to determine a transform that partitions the at least one non-connected segment into a first side and a second side opposite to the first side, wherein the first side includes a first 2D vertex and the second side include a second 2D vertex symmetric to the first 2D vertex; and
quantizing the first 2D vertex and the second 2D vertex in pairs to reduce a quantization error associated with the first 2D vertex and a symmetry prediction error associated with the second 2D vertex.
9. The method according to claim 8 , wherein the quantization error associated with the first 2D vertex is based on an error between the first 2D vertex and a quantization of the first 2D vertex.
10. The method according to claim 9 , wherein the quantization error associated with the first 2D vertex is further based on a difference between the second 2D vertex and a quantization of the second 2D vertex.
11. The method according to claim 10 , wherein the quantization of the second 2D vertex is determined based on a symmetry prediction performed on the quantized first 2D vertex and the transform.
12. The method according to claim 8 , wherein the quantizing the first 2D vertex and the second 2D vertex further comprises performing a search in a 2D space around the quantized first 2D vertex to find a best vertex on the first side of the at least one non-connected segment and a best vertex on the second side of the at least one non-connected segment that reduces the quantization error and the symmetry prediction error.
13. A encoder comprising:
at least one memory configured to store program code; and
at least one processor configured to read the program code and operate as instructed by the program code, the program code including:
performing code configured to cause the at least one processor to perform, on an input 3D mesh, a symmetry detection process to estimate a partition plane,
partitioning code configured to cause the at least one processor to partition the input 3D mesh into a first side and a second side based on the partition plane, wherein the first side is opposite to the second side, wherein a first vertex on the first side is symmetric to a second vertex on the second side; and
quantizing code configured to cause the at least one processor to quantize the first vertex and the second vertex in pairs to reduce a quantization error and symmetry prediction error associated with the first vertex and the second vertex.
14. The encoder according to claim 13 wherein the quantization error is based on an error between the first vertex and a quantization of the first vertex.
15. The encoder according to claim 14 , wherein the quantization error is further based on a difference between the second vertex and a quantization of the second vertex.
16. The encoder according to claim 15 , wherein the quantization of the second vertex is determined based on a symmetry prediction performed on the quantized first vertex and the partition plane.
17. The encoder according to claim 13 , wherein the quantizing further comprises performing a search in a 3D space around the quantized first vertex to find a best vertex on the first side and a best vertex on the second side that reduces the quantization error and the symmetry prediction error.
18. The encoder according to claim 13 , wherein the quantizing is performed on K symmetry pairs based on the partition plane, and
wherein the partition plane is updated based on an output of the quantizing performed on the K symmetry pairs.
19. The encoder according to claim 18 , wherein the quantizing performed on the K symmetry pairs and the updating of the partition plane is iteratively performed until a change in the partition plane between a first iteration and a second iteration is less than a threshold.
20. A non-transitory computer readable medium having instructions stored therein, which when executed by a processor in an encoder cause the processor to execute a method comprising:
performing, on an input 3D mesh, a symmetry detection process to estimate a partition plane;
partitioning the input 3D mesh into a first side and a second side based on the partition plane, wherein the first side is opposite to the second side, wherein a first vertex on the first side is symmetric to a second vertex on the second side; and
quantizing the first vertex and the second vertex in pairs to reduce a quantization error and symmetry prediction error associated with the first vertex and the second vertex.