熵編碼局部坐標(biāo)分級跳躍漸進式3D網(wǎng)格壓縮
發(fā)布時間:2018-01-27 15:44
本文關(guān)鍵詞: 貝葉斯熵編碼 局部坐標(biāo) 分級跳躍 網(wǎng)格壓縮 漸進式 高斯概率模型 邊沿觸發(fā) 出處:《計算機應(yīng)用研究》2017年10期 論文類型:期刊論文
【摘要】:為進一步提高三維網(wǎng)格壓縮算法性能,在高斯混合概率模型(GHPM)基礎(chǔ)上,提出基于貝葉斯熵編碼的局部坐標(biāo)分級跳躍漸進式3D網(wǎng)格壓縮算法。采用GHPM模型實現(xiàn)3D網(wǎng)格壓縮過程的頂點創(chuàng)建、邊沿觸發(fā)器設(shè)計、面方向預(yù)測以及分級跳躍分割,實現(xiàn)對給定頂點的后驗概率幾何拓?fù)浞柟烙。基于后驗概率的算術(shù)編碼器進行拓?fù)浞柧幋a,采用不同情景進行設(shè)計,提出漸進式的標(biāo)簽預(yù)測過程,實現(xiàn)已編碼組信息的充分利用,并采用局部坐標(biāo)系有效壓縮幾何殘差。通過與對比編碼器的實驗驗證,所提算法相對于AD、wavemesh、AAD以及RDO編碼器具有更高的壓縮比和壓縮精度,計算性能更好。
[Abstract]:In order to further improve the performance of 3D mesh compression algorithm, Gao Si hybrid probability model based on GHPM-based. A progressive 3D mesh compression algorithm based on Bayesian Entropy coding is proposed. The vertex creation and edge trigger design of 3D mesh compression process are realized by using GHPM model. The geometric topological symbol estimation of a given vertex is realized by prediction of surface direction and hierarchical jump segmentation. The arithmetic encoder based on posteriori probability encodes topological symbols and designs with different scenarios. A progressive label prediction process is proposed to make full use of the coded group information, and the geometric residuals are effectively compressed using local coordinate system. The experimental results show that the proposed algorithm is relative to AD. The Wavemes AAD and RDO encoders have higher compression ratio and compression accuracy and better computational performance.
【作者單位】: 河南工學(xué)院計算機科學(xué)與技術(shù)系;河南師范大學(xué)計算機與信息工程學(xué)院;
【基金】:國家自然科學(xué)基金資助項目(U1404602) 河南省高等學(xué)校重點科研項目(15B520006,15A520063) 河南省教育廳科學(xué)技術(shù)研究重點項目(14A520046)
【分類號】:TP393.02
【正文快照】: 0引言隨著圖形和互聯(lián)網(wǎng)技術(shù)進步,3D網(wǎng)格壓縮成為實現(xiàn)高效網(wǎng)格模型存儲和傳輸?shù)年P(guān)鍵。三角形網(wǎng)格包含幾何數(shù)據(jù)拓?fù)浣Y(jié)構(gòu)。連通性數(shù)據(jù)用于區(qū)分拓?fù)浣Y(jié)構(gòu)或頂點間的連通信息,而幾何數(shù)據(jù)描述頂點位置。根據(jù)解碼策略,3D網(wǎng)格壓縮可分為單速率和漸進式編碼器。前者可讀取完整比特流實現(xiàn),
本文編號:1468818
本文鏈接:http://sikaile.net/guanlilunwen/ydhl/1468818.html
最近更新
教材專著