US 12,307,725 B2
Point cloud intra prediction method and device based on weights optimization of neighbors
Ge Li, Guangdong (CN); Qi Zhang, Guangdong (CN); Yiting Shao, Guangdong (CN); and Jing Wang, Guangdong (CN)
Assigned to Peking University Shenzhen Graduate School, (CN)
Appl. No. 17/631,333
Filed by Peking University Shenzhen Graduate School, Guangdong (CN)
PCT Filed Sep. 11, 2019, PCT No. PCT/CN2019/105308
§ 371(c)(1), (2) Date Jul. 12, 2022,
PCT Pub. No. WO2021/022621, PCT Pub. Date Feb. 11, 2021.
Claims priority of application No. 201910717002.5 (CN), filed on Aug. 5, 2019.
Prior Publication US 2023/0196625 A1, Jun. 22, 2023
Int. Cl. G06K 9/00 (2022.01); G06T 9/00 (2006.01)
CPC G06T 9/001 (2013.01) 7 Claims
OG exemplary drawing
 
1. A point cloud intra prediction method based on weight optimization of a neighbor, comprising:
determining K nearest neighboring points of a current point;
calculating an optimized weight of each nearest neighboring point of the current point according to a coordinate of the current point and coordinates of the K nearest neighboring points by optimizing corresponding coefficients of x, y and z coordinate components of distances between the current point and the K nearest neighboring point;
calculating the differences between the current point and the nearest neighboring point on the x, y and z coordinate components and calculating the square values respectively, so as to obtain quadratic differences between the current point and the nearest neighboring point on the x, y and z coordinate components;
respectively multiplying the quadratic differences between the current point and the nearest neighboring point on the x, y and z coordinate components by corresponding coefficients α, β and γ, so as to obtain weighted quadratic differences between the current point and the nearest neighboring point on the x, y and z coordinate components;
summing the three weighted quadratic differences between the current point and the nearest neighboring point on the x, y and z coordinate components, so as to obtain a weighted distance square value between the current point and the nearest neighboring point;
calculating a reciprocal value of the weighted distance square value between the current point and the nearest neighboring point, so as to obtain the optimized weight of the nearest neighboring point;
carrying out weighted summation on attribute reconstruction values of the K nearest neighboring points by utilizing the optimized weights of the K nearest neighboring points of the current point, so as to obtain an attribute prediction value of the current point; and
carrying out coding processing according to the attribute prediction value of the current point, comprising:
determining a prediction residual of the current point by calculating the difference between an attribute value and the attribute prediction value of the current point; and encoding the prediction residual by carrying out transformation, quantization and entropy coding, so as to obtain a bitstream.