Point cloud processing method, decoder and storage medium

A point cloud processing method for improving point cloud resolution based on a generative adversarial network (GAN), wherein the GAN comprises a generator, wherein the generator comprises a feature extraction network, a coarse feature expansion network, a geometry generation network and a geometry offset prediction network, the method is performed by a computer device including a processor, and the method comprises:

• inputting, through the processor, a first point cloud into the feature extraction network to obtain a first feature of a point of the first point cloud;
• inputting, through the processor, the first feature into the coarse feature expansion network to perform up-sampling on the first feature to obtain a second feature;
• inputting, through the processor, the second feature into the geometry generation network to map the second feature from a feature space to a geometry space to obtain a second point cloud;
• inputting, through the processor, the second feature and the second point cloud into the geometry offset prediction network to determine an offset value of a point in the second point cloud in the geometry space, the offset value being used for correcting the second point cloud;
• and
• obtaining, through the processor, a third point cloud based on the second point cloud and the offset value, the third point cloud having a higher resolution than the first point cloud,
• wherein the GAN further comprises a discriminator, and the method further comprises:
• inputting, through the processor, the third point cloud into the discriminator to obtain a confidence value of a point in the third point cloud, wherein the confidence value indicates whether the third point cloud is ground truth,
• wherein inputting, through the processor, the third point cloud into the discriminator to obtain the confidence value of the point in the third point cloud comprises:
• inputting, through the processor, the third point cloud into at least two consecutive second multilayer perceptron (MLP) layers;
• performing, through the processor, a maximum pool operation on a feature output by a last second MLP layer in the at least two consecutive second MLP layers to obtain a global feature vector of a third feature;
• concatenating, through the processor, the global feature vector and features output by the at least two consecutive second MLP layers to determine a concatenated feature;
• and
• inputting, through the processor, the concatenated feature into a third autocorrelation attention mechanism, and inputting a result output by the third autocorrelation attention mechanism into a second full connection (FC) layer to obtain the confidence value.