Neural gas based 3D normal mesh compression

The recent widespread of processing and transmitting 3D model in various fields such as computer graphics, animations and visualization calls an essential need for efficient geometry mesh compression technique that became more crucial. This paper explores a progressive compression technique for 3D normal meshes geometry by utilizing one of competitive learning methods. The introduced technique is based on multi-resolution decomposition which was obtained by wavelet transformation. Then the coefficients are quantized by neural gas algorithm as a vector quantizer which improves the visual quality of the reconstructed geometry mesh. Our experiments show that the explored technique out performs the state-of-art techniques in Terms of visual quality of compressed meshes.