ZTE Inotungamirwa Algorithm yeLossless Point Cloud Geometry Compression

Zvinodiwa:

• Product Name: Spatio-Temporal Context-Guided Algorithm for Lossless Point Cloud Geometry Compression
• Vanyori: ZHANG Huiran, DONG Zhen, WANG Mingsheng
• Rakatsikiswa: Zvita 2023
• DOI: 10.12142/ZTECOM.202304003

Mirayiridzo Yekushandiswa Kwechigadzirwa

Nhanganyaya:
Chigadzirwa chacho chakagadzirirwa kunyatso compress point Cloud data, kugadzirisa matambudziko ane chekuita nenzvimbo yekuchengetedza uye network yekufambisa bandwidth.

Main Features:

1. Prediction mode inoshanda kune intraframe uye inter-frame point makore uchishandisa iyo yakawedzera yekufamba mutengesi dambudziko.
2. Adaptive arithmetic encoder ine kukurumidza mamiriro ekugadzirisa kuti igone kuverenga uye kumisikidza mhinduro.

Matanho Ekushandisa:

Danho 1: Kamura Point Makore
Kamura mapoinzi makore kuita mayuniti layers pamwe chete neakisi huru.

Danho 2: Dhizaina Prediction Mode
Dhizaina maitiro ekufungidzira uchishandisa algorithm yekufamba mutengesi kukwidziridza nzvimbo uye yenguva diki diki.

Danho rechitatu: Encode Residuals
Nyora zvakasara kuita bit streams uchishandisa context-adaptive arithmetic encoder yekumanikidza.

FAQ:

• Mubvunzo: Ndezvipi zvakakosha zvekushandisa chigadzirwa ichi?
  A: Chigadzirwa chinogonesa kudzvanya kwakanaka kweiyo point Cloud data, ichishandisa spatial uye temporal corelations kune yakawedzera compression mhinduro.
• Mubvunzo: Chigadzirwa ichi chinogona kubata ese ari maviri-frame uye akawanda-frame point makore?
  A: Ehe, iyo yekufembera modhi inoshanda kune ese ari maviri intraframe uye inter-frame point makore, achibvumira akasiyana-siyana ekushandisa mamiriro.

ZHANG Huiran, DONG Zhen, WANG Mingsheng

1. Guangzhou Urban Planning uye Design Survey Research Institute, Guangzhou 510060, China;
2. Guangdong Enterprise Key Laboratory yeUrban Sensing, Monitor⁃ ing uye Yambiro Yepakutanga, Guangzhou 510060, China;
3. State Key Laboratory yeRuzivo Injiniya muKuongorora Mepu⁃ ping uye Remote Sensing, Wuhan University, Wuhan 430079, China)

Abstract: Point Cloud compression yakakosha kuendesa 3D inomiririra yepasirese senge 3D inonyudza telepresence, autonomous kutyaira, uye tsika heri.tage kuchengetedza. Nekudaro, point Cloud data inogovaniswa zvisina tsarukano uye nekusaenderera munzvimbo dzepasi uye dzenguva, uko kusinga shandiswe ma voxels uye isina kusimba kuwirirana munzvimbo ye 3D kunoita kuti kubudirira kudzvinyirira kuve dambudziko rakaoma. Mubepa rino, isu tinokurudzira spatio-temporal mamiriro-inotungamirwa algorithm yekusarasikirwa point Cloud geometry compression. Chirongwa chakarongwa chinotanga nekupatsanura gore repoindi kuita zvidimbu zvakachekwa zveukobvu hweyuniti pamwe neakisi refu refu. Zvadaro, inosuma nzira yekufanotaura uko zvese intraframe uye inter-frame point makore aripo, nekuona matsamba pakati pematanho ari pedyo uye kufungidzira nzira ipfupi uchishandisa algorithm yekufamba mutengesi. Chekupedzisira, mashoma ekufanotaura akasara akanyatso kudzvanywa neyakakwana-inotungamirwa uye inogadzirisa fastmode arithmetic coding matekiniki. Maedzo anoratidza kuti nzira yakatsanangurwa inogona kubudirira kudzika yakaderera mwero isingarasikike yekumisikidzwa kweiyo point Cloud geometric ruzivo, uye inokodzera 3D point Cloud compression inoshanda kune akasiyana marudzi ezviratidziro.
Keywords: point cloud geometry compression; single-frame point makore; multi-frame point makore; predictive coding; arithmetic coding.

Citation (Format 1): ZHANG HR, DONG Z, WANG M S. Spatio-temporal context-guided algorithm for lossless point cloud geometry compression [J]. ZTE Communications, 2023, 21(4): 17–28. DOI: 10.12142/ZTECOM.202304003
Citation (Format 2): HR Zhang, Z. Dong, naMS Wang,"Spatio-temporal mamiriro-inotungamirwa algorithm yekusarasikirwa point Cloud geometry compression," ZTE Communications, vol. 21, kwete. 4, mapeji 17–28, Zvita 2023. doi: 10.12142/ZTECOM.202304003.

 

Nhanganyaya

Nekuvandudzwa kwema-multi-platform uye akawanda-resolution acquisition equipment performance, chiedza chekuona uye kubva (LiDAR) tekinoroji inogona kunyatso kutevedzera zvinhu zve3D kana zviono zvine hombe mapoinzi seti. Kuenzaniswa neyechinyakare multimedia data, point Cloud data ine ruzivo rwekuyera rwemuviri runomiririra zvinhu kubva mahara viewmapoinzi, kunyangwe zviratidziro zvine maumbirwo akaomarara epamusoro. Izvi zvinoguma nekupindirana kwakasimba uye kunyudza mhedzisiro inopa vashandisi ruzivo rwakajeka uye rwechokwadi rwekuona. Pamusoro pezvo, point Cloud data ine yakasimba anti-ruzha kugona uye parallel kugadzirisa kugona, izvo zvinoita kunge zvakawana kukwezva kubva kuindasitiri uye academia, kunyanya kune nzvimbo dzekushandisa senge tsika heri.tage kuchengetedza, 3D immersive telepresence uye kutyaira otomatiki [1-2].
Nekudaro, point Cloud data kazhinji ine mamirioni kusvika kumabhiriyoni emapoinzi munzvimbo dzepasi, zvichiunza mitoro nematambudziko kune yekuchengetera nzvimbo kugona uye network yekufambisa bandwidth. Semuyenzaniso, yakajairika dynamic point gore rinoshandiswa kuvaraidzo rinowanzo sanganisira miriyoni imwe yemapoinzi pafuremu, iyo, pamafuremu makumi matatu pasekondi, inokwana bandwidth yakazara ye30 Gbit/s kana ikasiiwa isina kumisikidzwa[3.6]. Naizvozvo, tsvagiridzo yepamusoro-inoshanda geometry compression algorithms yenzvimbo yemakore ine yakakosha theoretical uye inoshanda kukosha.
Basa risati raitwa rakagadzirisa dambudziko iri nekuvaka zvakananga magridi kana pa-kuda pasi-sampling, nekuda kwekuganhurirwa musimba rekombuta yekombuta uye point Cloud kuunganidza kushanda zvakanaka, izvo zvakakonzera kuderera kwespatio-temporal compression performance uye kurasikirwa kweiyo geometric hunhu hweruzivo ruzivo. Zvidzidzo zvenguva pfupi yapfuura zvaive zvakanyanya kubva pamakomputa magiraidhi uye madhijitari masaini ekugadzirisa maitiro ekushandisa block mashandiro pa-point cloud data [4 5] kana yakasanganiswa vhidhiyo yekukodha tekinoroji [6 7] yekugonesa. Muna 2017, iyo Moving Picture Experts Group (MPEG) yakakumbira zvikumbiro zvekutsikirira kwegore uye yakaita nhaurirano dzakazotevera dzekumanikidza rudzi urwu rwe data. Nekuwedzera kwemaitiro ekunongedza cloud comression iripo uye kuratidzwa, maviri-point cloud data compression frameworks-TMC13 neTMC2 akabudiswa muna 2018. Tsvakurudzo iri pamusoro inoratidza kufambira mberi kunoshamisa kwakaitwa mutekinoroji yekumanikidza yepoint cloud. Zvakadaro, basa rekutanga rainyanya kubata nepakati nepakati pehukama hwemapfuti makore zvakaparadzana asi rakanga risati rashandiswa mukugona kwavo kuzere mukutsikirira kwegore.
Kuti tigadzirise matambudziko ambotaurwa, tinosuma spatio-temporal mamiriro-inotungamirwa nzira yekusarasikirwa point Cloud geometry compression. Isu tinotanga kupatsanura mapoinzi makore kuita zvikamu zveyuniti padivi peiyo main axis. Isu tinobva tagadzira maitiro ekufungidzira kuburikidza neanofamba mutengesi algorithm, nekutora spatiotemporal correlation. Chekupedzisira, izvo zvakasara zvinonyorwa kuita bit-nzizi ine yakashandiswa mamiriro-inogadzirisa arithmetic encoder. Mipiro yedu huru ndeiyi inotevera.
1) Isu tinogadzira- maitiro ekufanotaura anoshanda kune ese ari mukati-frame uye inter-frame point gore, kuburikidza neakawedzera anofamba mutengesi dambudziko (TSP). Nekusimudzira zvese zvenzvimbo uye zvenguva diki dzemakore, kufanotaura kwejometri kunogona kushandisa zvirinani kuwirirana kwenzvimbo uye nekudaro kugonesa marudzi akasiyana ezviitiko.
2) Isu tinopa inogadzirisa arithmetic encoder ine nekukurumidza con-text update, iyo inosarudza iyo yakakwana 3D mamiriro kubva muduramazwi rechirevo, uye inodzvanyirira kuwedzera kwekufungidzira kwe entropy. Nekuda kweizvozvo, inosimudzira mukana wekuverenga kugona kweentropy encoder uye inopa yakakosha mhedzisiro.
Imwe yese bepa iri yakarongwa sezvizvi. Chikamu chechipiri chinopa rondedzero yebasa rakabatana pane point Cloud geometry compression. Chikamu 2 chekutanga chinopa pamusoroview yegadziriro yakarongwa. Zvadaro, nzira yakarongwa inotsanangurwa zvakadzama. Mibairo yekuedza nemhedziso inoratidzwa muChikamu 4 ne5, zvichiteerana.

Basa Rinoenderana

Pave paine akawanda point Cloud geometry compression algorithms yakakurudzirwa mumabhuku. CAO et al. [8] uye GRAZIOSI et al. [9] ita ongororo uye pfupiso yeazvino point cloud compression nzira, inotarisa pane spatial dimension compression tekinoroji uye MPEG yekumisikidza masisitimu zvichiteerana. Isu tinopa pfupiso review yezvichangobva kuitika muzvikamu zviviri: single-frame point cloud compression uye multi-frame point cloud compression.

1. Single-Frame Point Cloud Compression
   Single-frame point makore anoshandiswa zvakanyanya muinjiniya ongororo, tsika heritage kuchengetedza, geographic information system, uye zvimwe zviitiko. Iyo octree inzvimbo inoshandiswa zvakanyanya data data kuti imiririre zvakaringana makore, ayo anogona kumanikidzwa nekurekodha ruzivo kuburikidza neakagarwa node. HUANG et al.[10] kurudzira nzira yeoctree-yakavakirwa iyo inodzokorodza kupatsanura gore repfundo mumanodhi nenzvimbo dzadzo dzinomiririrwa nepakati pejometri yeyuniti yega yega. FAN nevamwe.[11] wedzera kunatsiridza nzira iyi nekuunza cluster analysis kuti ibudise nhanho yeruzivo (LOD) hierarchy uye nekuiisa muhupamhi-yekutanga kurongeka. Zvisinei, nzira idzi dzinogona kukonzera kukanganisa nekuda kwekufungidzira kwemuenzaniso wepakutanga panguva yekudzokorora.
   Kugadzirisa izvi zvisingakwanisi, nyanzvi dzakaunza maumbirwo ejometri, zvakadai se- triangular surface model [12], planar surface model [13 14], uye clustering al-gorithm [15], yekufungidzira kwepakati-kati uye kusara kwekuverenga. . RENTE nevamwe.[16] kurudzira pfungwa yekufambira mberi yakaturikidzana iyo inotanga kushandisa iyo octree chimiro kune yakakasharara-grained encoding uye yobva yashandisa girafu yekushandura Fourier yekutsikirira uye kuvaka patsva yemakore. Muna 2019, MPEG yakaburitsa tekinoroji yejometry-based point cloud compression (G-PCC) tekinoroji yemakore ese akamira uye ane simba, ayo anoitwa kuburikidza nekubatanidza shanduko, voxelization, geometric chimiro chekuongorora, uye arithmetic coding nhanho nhanho[17].
   Sezvo mamwe maoctants mukati meoctree anogona kunge aine vanhu vashoma kana asina chinhu, dzimwe nzira dzakataurwa kuti dzigadzirise chimiro chemuti nekuchekerera sub-node uye nekudaro kuchengetedza ndangariro kugoverwa. For example, DRICOT et al. [18] kurudzira inferred yakananga coding mode (IDCM) yekumisa iyo octree partition zvichibva pane zvakafanotsanangurwa mamiriro e sparsity kuongororwa, izvo zvinosanganisira kuchekerera iyo octree chimiro kuchengetedza mabits akagoverwa kune vana node. ZHANG et al. [19] inokurudzira kupatsanura nzvimbo yegore pamwe chete nezvikamu zvikuru uye kugadzirisa nzira yekuparadzanisa kubva pamuti webhinari, quadtree uye octree. Kuenzaniswa neyechinyakare octree partitioning, iwo mahybrid modhi ataurwa pamusoro anogona kunyatso kudzikisa huwandu hwemabhiti anoshandiswa kumiririra sparse mapoinzi, nekudaro kuchengetedza node dzinoda kuvharwa. Nekudaro, yakaoma hyperparameter mamiriro uye kugadzwa kwemodhi inodiwa mukuita, zvichiita kuti zviome kuzadzisa zvinodiwa zvekuzvigadzirisa uye kuderera kuoma.
   Neyakadzika neural network ichiita nhanho dzakakura mumufananidzo uye vhidhiyo compression, vaongorori vakatsvaga nzira dzekuwedzera kudzikisira mareti nekusimudzira super pamberi pekutungamira uye kudzokororwa kwelatent nzvimbo yekutaura panguva yekumanikidza. QUACH nevamwe.[20] uye HUANG et al. [21] kurudzira nzira dzinobatanidza pfungwa idzi. GUARDA et al. sanganisa convolutional neural network uye autoencoders kushandisa redundancy pakati penzvimbo dziri padyo uye kuwedzera coding kuchinjika muRef. [22]. Munguva pfupi yapfuura, WANG et al. [23] propose point Cloud compression nzira yakavakirwa pane ariational auto-encoder, iyo inovandudza iyo compression reshiyo nekudzidza hyperprior uye kuderedza kurangarira kushandiswa kwemasvomhu ekodha. Idzi nzira dzambotaurwa dzinoshandisa neural network encoder kubata iyo yepamusoro-yakavanzika vector yepoint cloud, iyo entropy modhi zvingangoitika, uye mukana wemupendero unokwana zvirinani, nokudaro kuderedza kushandiswa kwendangariro kwemasvomhu ekodha. Kazhinji kutaura, tsvagiridzo yeimwe-frame point Cloud geometric compression ine hukuru, asi pane matambudziko maviri asara. Spatial correlation haina kushandiswa zvinobudirira, uye nzira zhinji hadzina kodhi kuwirirana kweiyo point Cloud data zvakakwana uye nemazvo. Kunze kwezvo, kuverenga kweiyo mukana wemodhi ye entropy coding inoratidzika kureba uye yakaoma nekuda kwehuwandu hukuru hwemamiriro.
2. Multi-Frame Point Cloud Compression
   Multi-frame point makore anowanzo shandiswa muzviitiko zvakaita senge chaiyo-nguva 3D inonyudza telepresence, inopindirana VR, 3D yemahara. viewpoint nhepfenyuro uye otomatiki kutyaira. Kusiyana ne-single-frame point cloud compression, multi-frame point cloud compression inokoshesa kushandiswa kwekugadzirisa nguva, pamwe chete nekufungidzira kwekufamba uye muripo. Nzira dziripo dzeakawanda-frame point Cloud compression inogona kukamurwa kuita mapoka maviri: 2D fungidziro uye 3D kushongedza.
   Munda wekudzvanya kwemufananidzo uye vhidhiyo wakakura uye wakanyatsoongororwa mumakumi mashoma emakore apfuura. Akasiyana-siyana maalgorithms anoshandura mapoint cloud kuita mifananidzo yobva yaamanikidzira zvakananga neFFmpeg neH. 265 encoder, etc. AINALA et al[24] vanounza planar proximate approximate en-coding mode iyo inokodha zvese geometry uye ruvara hunhu kuburikidza neraster scanning mundege. . Nekudaro, iyi nzira inokonzeresa shanduko muchimiro chakanangwa panguva yekugadzira mepu, zvichiita kuti kupindirana-kufanotaura kuve kwakaoma. Naizvozvo, SCHWARZ et al.[25] uye SEVOM et al. [26] inokurudzira yakatenderedzwa planar fungidziro, cube fungidziro, uye chigamba-yakavakirwa fungidziro nzira dzekushandura mapoinzi makore kuita mavhidhiyo e2D, zvichiteerana. Nekuisa mafungidziro akafanana mumafuremu ari padyo panzvimbo imwechete mumifananidzo iri padyo, vhidhiyo compressor inogona kubvisa zvizere kuwirirana kwechinguvana. MuRef. [27], kufembera kwepakati-geometry kunoitwa kuburikidza neTSP, iyo inoverengera iyo imwe-kune-imwe tsamba yepedyo intra-blocks nekutsvaga bhuroka ine yepedyo avhareji kukosha. MPEG yakaburitsa vhidhiyo-based point Cloud compression (V-PCC) tekinoroji yeakasimba point makore muna 2019[28]. Iyi dhizaini inokamura gore rekupinza kuita mabhururu madiki ane akafanana akajairwa mavheji uye inoenderera nzvimbo, wozoashandura kune iyo planar pamusoro kuburikidza nemacubes kurekodha mufananidzo wekugara uye ruzivo rwekubatsira. Yese mifananidzo inoguma inomanikidzwa neakura vhidhiyo macodecs, uye ese bitstreams anounganidzwa kuita imwechete inobuda file. Kumwe kuedza kwakaitwa kuvandudza kushanda kwenzira idzi. COSTA nevamwe. [29] shandisa akati wandei matsva chigamba ekurongedza nzira kubva pakuona kwekugadzirisa kweiyo yekurongedza algorithm, data packaging link, zvine hukama kuronga, uye zviratidziro zvekumisikidza. Uyezve, PARK et al. [30] gadzira data-adaptive packing nzira iyo inogadzirisa mapoka ari padyo nemafuremu muboka rimwechete zvinoenderana nechimiro chakafanana pasina kukanganisa mashandiro eV-PCC rukova. Nekuda kwekurasikirwa kweruzivo kusingadzivisike kunokonzerwa neiyo point fungidziro yegore, nyanzvi dzakagadzira nzira dzinoshanda dzekudzvanya iyo point Cloud kutevedzana kwemafuremu anotevedzana vachishandisa tekinoroji yemuripo wakavakirwa panzvimbo ye3D. KAMMERL et al.[31] kurudzira octree-yakavakirwa geometric encoding nzira, iyo inogonesa yakakwirira yekumanikidza nekuita iyo yakasarudzika OR (XOR) misiyano pakati pemafuremu ari padyo. Iyi nzira haina kungogamuchirwa muPoint Cloud Library yakakurumbira (PCL) [32] asi yakashandiswa zvakanyanya kutsvagisa algorithm. Dzimwe nzira dzeinterframe dzinoshandura dambudziko refungidziro yekufamba kwe3D kuita dambudziko rinoenderana [33] kana kushandisa ruzivo rwejometri yakagadziridzwa[34] kufanotaura mafambiro uye kuona hukama hunoenderana pakati pemafuremu ari pedyo nemazvo. Zvidzidzo zvinoputika [35 36] zvakaratidza kuti kudzvanywa kwevhidhiyo kwakadzidzwa kunopa zvirinani-kukanganisa kuita pane zvechinyakare, zvichiunza kukosha kwakakosha kunongedza kudzvanywa kwegore. ZHAO nevamwe. [37] suma bi-directional inter-frame prediction network kuti iite kufanotaura kweinter-frame uye kuunza mashandisirwo akakodzera eruzivo rwakakodzera mune zvepasi uye zvenguva. KAYA et al. [38] gadzira paradigm nyowani yekuisa encoding geometric maficha edense point Cloud sequences, kugadzirisa iyo CNN yekufungidzira kugovera encoding kuona kusarasikirwa kwekumanikidzana kwemakore akakora.
   Kunyangwe kufambira mberi muiyo compression coding tekinoroji yeakawanda-frame point cloud modhi, matambudziko maviri anoenderera mberi. Iyo iripo yakawanda-frame point Cloud compression maitiro anonyanya kuvimba nekodha yevhidhiyo uye muripo wekufamba, izvo zvinosanganisira kurasikirwa neruzivo kana kukanganisa kunokonzerwa nemepu uye kuvharira kumucheto kusaenderera. Pamusoro pezvo, kufanotaura kodhi kunoratidza kuderera kwekushandisa nekuda kwekusaenderana kweiyo inter-frame point cloud geometry. Iko kuratidzika kubviswa kwemapoinzi pakati pemafuremu uye ruzha rusingadzivisike kunowedzera kuoma kwekushandisa zvinobudirira kufungidzira coding mukutsikirira kwe-inter-frame.

Proposed Spatio-Temporal Context-Inotungamirwa Lossless Geometry Point Cloud Compression Method

Overview
Iyo pombi yakazara yespatio-temporal inotungamirwa algorithm inoratidzwa muFig. 1. Chekutanga, isu tinofanogadzirisa iyo yekupinda point Cloud nekushandisa voxelization uye chiyero shanduko. Zvadaro, gore repoindi rakakamurwa kuita yuniti-ukobvu akachekwa zvidimbu pamwe chete neakisi huru. Tevere, tinogadzira nzira yekufanotaura iyo inoshandisa zvizere ruzivo rwenguva uye spatial correlation mukati mese intra-frame uye inter-frame. Isu tinoverenga iyo ipfupi nzira yemapoinzi ereferenzi layer (R-layers) kuburikidza neanofamba mutengesi algorithms, uye mibairo yeR-layers inozoshandiswa kufanotaura spatiotemporally uye encode mamwe makore ese epoinzi, ayo akafanotaurwa layers (P-layers. ) Pakupedzisira, iyo yakagadziridzwa entropy coding algorithms inogamuchirwa kuti iwane yakamanikidzwa bhinari file.

Image Sliced-Based Hierarchical Division

1. Pre-processing
   Iyo pre-processing module inosanganisira voxelization uye chiyero shanduko, kuitira zvirinani indexing yeimwe neimwe pfungwa. Mune voxelization, tinogovanisa nzvimbo kuita ma cubes ehukuru N, iyo inoenderana nekugadziriswa chaiko kweiyo point cloud. Imwe neimwe poindi inopihwa yakasarudzika voxel zvichienderana nenzvimbo yayo. Voxel yakanyorwa se1; kana yakagarwa zvakanaka, iri 0 neimwe nzira. Shanduko yechikero inogona kuderedza sparsity yekumanikidza zvirinani nekubvisa iyo point gore, uko chinhambwe pakati pemapoinzi chinova chidiki. Isu tinounganidza iyo point Cloud coordinates (x, y, z) tichishandisa kuyera factor s, kureva,
   Kuti tive nechokwadi chekusarasikirwa kwekumanikidzana, isu tinofanirwa kuve nechokwadi chekuti scaling factor s haigone kukonzera kurasikirwa kwejometri uye inoda kudhindwa zvakare mumusoro. file.
2. Sliced-layer division
   Iyi module inoshanda nekukamura iyo 3D point gore pamwe chete nedemo rayo, ichigadzira akati wandei mayuniti-akachekwa akaturikidzana ane akagarwa uye asina-akagarwa ruzivo chete anogona kuwedzeredzwa nekushandisa encoder yekufungidzira uye arithmetic coder. Basa racho rinotsanangurwa se:
   apo G inoreva nzvimbo yekuisa gore rinobatanidza matrix, axis inoreva chiyero chakasarudzwa, uye S (a, b) ndiyo 2D chimedu chinobviswa nechero yega. Kazhinji, isu tinoitisa zviedzo pahuwandu hukuru hwekutevedzana kwebvunzo, uye mhedzisiro inoratidza kuti kupatsanura pamwe neakisi refu yepoint cloud spatial variation inopa yakaderera bitrate, kureva.
3. Minimum bounding box kubviswa
   Muzviitiko zvakawanda, ma-voxels anogarwa kazhinji haagone kudziviswa uye anodarika zvakanyanya mavoxel anogarwa. Nekuda kweizvozvo, kugadzirisa uye encoding ese ari maviri marudzi evoxels panguva imwe chete inoremedza iyo computational kuomarara uye encoding kumhanya kweiyo compression algorithm. Nokudaro, tinotora bhokisi rinosungirirwa (OBB) [39] kuverenga bhokisi rinosungirirwa rinosungirirwa kune rimwe nerimwe rakachekwa, kuve nechokwadi chokuti mitsara yemabhokisi anosungirirwa inopindirana kumativi ose. Mukugadzirisa kunotevera, chete ma voxels ari mukati merectangle yakaganhurirwa anomanikidzwa.

Spatial Context-Guided Predictive Encoding

Chinangwa chenzvimbo-inotungamirwa yekufungidzira encoding ndeye encode ese mapoinzi layer by layer. Kufemerwa neTSP, isu tinogadzira maitiro ekufanotaura kuti tiongorore zvingangoita maodha uye kuwirirana mukati mega ega akachekwa layer. Iyi module ine chikamu uye ipfupi nzira kuverenga.
Pakutanga, tinoparadzanisa zvidimbu zvakachekwa uye tinoona R-layer uye R-layers yeboka rimwe nerimwe. Isu tinoyambuka point cloud layer ne layer pamwe neaxis yakasarudzwa. Kana kureba kwegwaro guru rebhokisi rinosungirirwa diki pakati pezvikamu zviri pedyo zvichisiyana nehurefu hweyuniti yakatarwa, inorekodhwa seboka rimwe chete. Zvikasadaro, rinoshandiswa sereferensi yeboka rinotevera, uye yega yega gore muboka rinotevera rinoshandisa nzira imwechete ipfupi. Mupepa iri, tinoisa chikamu chekutanga cheboka rimwe nerimwe seR-layer, uye mamwe seP-layers. Isu zvakare tinoita zviedzo pahuwandu hukuru hweyedzo se-uence uye tinokurudzira kuseta iyi yakatsanangurwa parameter semayuniti matatu kuti uwane yakanyanya kudzvanya.
Mushure mezvo, isu tinoitisa iyo ipfupi nzira yekuverenga paR-layers uye kurekodha zvakasara zveVatambi. Zvinoenderana nemutemo wekugovera wepoindi yegore reyega yega chidimbu, isu tinorongedzera zvisingaite makore mapoinzi ega ega dhizaini zvichienderana neTSP algorithm. Izvi zvinotitendera kuti tinyatso kuverengera nzira ipfupi kune iyo point cloud yeR-layers, tobva tanyora zvakasara zvezvikamu zvinofanotaura. Algorithm 1 inoratidza pseudo-code yemaitiro ekufanotaura.

Chekutanga, tinotsanangudza mutemo wekuverenga chinhambwe pakati pemapoinzi munzvimbo yenzvimbo uye tanga nzira yegwara nechero yakasarudzwa yakasarudzwa pc1. Mukudzokorora kwega kwega, pese panowedzerwa poindi nyowani pci, mvumo yacho inovandudzwa zvine simba kuburikidza neiyo state transition equation nzira (P - i, i) kusvika mapoinzi ese akawedzerwa akanyorwa muP mukurongeka kwenzira ipfupi. Iyi nzira inogadziridzwa zvishoma nezvishoma zvichienderana nediki diki danho. Mushure mekunge kudzokororwa kwese kwapedzwa munzira yakapfupika, tinoverenga min dist(pci, pcj) mune imwe neimwe yeR-layers, uye dzosera iyo ipfupi nzira rekodhi tafura yemapoint makore mune imwe neimwe yeR-layers. Kuti uwedzere kudzvinyirira, tinoverenga kutsauka kweP-layers kubva munzira pfupi yeR-layer mukati meboka rimwechete uye tinonyora sezvisaririra zvekufungidzira. Chekupedzisira, nzira ipfupi yeRlayer uye zvakasara zveboka rega rega zvinobuda uye zvinopfuudzwa kune entropy encoder kudzvanya zvisaririra zvekufungidzira.

Spatio-Temporal Context-Guided Predictive Encoding
The spatial inotungamirwa nemamiriro ekufanotaura modhidhi inoisa
single-frame point makore ari oga. Nekudaro, kushandisa encoding yenzvimbo kune yega yega-frame point Cloud zvakaparadzana zvinogona kupotsa mikana yakafumurwa neiyo temporal corelations pamulti-frame point cloud. Tichifunga nezve iyo yakawanda-frame point gore inogovanisa machunks akakura ekupfuura, isu tinotarisa pakushandisa temporal redundancy kuti tiwedzere kusimudzira kushanda nesimba. Nekudaro, zvichibva pane yakarongwa yenzvimbo-inotungamirwa yekufungidzira modhi, tinokwanisa kudzvanya multiframe point gore nekuona tsamba inonyorerana pakati pezvikamu zviri padhuze pamafuremu.

1. Inter-frame partition
   Kuti uwedzere kushanda kweinter-frame kufanotaura modhi, zvakakosha kuti uve nechokwadi chekufanana kwakaringana pakati peadja⁃cent layers emafuremu. Nekuda kweizvozvo, isu tinofanirwa kupatsanura mapoka pakati pemafuremu ari padyo uye toona maR-layers neP-mafuremu. Nekufungidzira nzira ipfupi yeP-layers zvichibva panzira ipfupi yeR-layers, tinorekodha zvakasara zvekufungidzira uye tozvimanikidza kuburikidza neiyo entropy encoder. Algorithm 2 inoratidza iyo pseudocode yeiyo interframe partition.
   Zvichienderana neakachekwa-matanho ekumisikidzwa, isu tinoona yakakasharara kupatsanura uye yakanaka partition zvakatevedzana. Kune yakaomesesa partition, isu tinoronga akachekwa akaturikidzana ega ega furemu zvichienderana nemakonisheni anoenderana nemasanhu ekukamura, kubva kudiki kusvika kuhombe. Nekuda kweizvozvo, chidimbu chega chega chefuremu yega yega chine nhamba yakasarudzika, zvichitibvumira kugovanisa zvimedu zvimedu nenhamba yakafanana pakati pemafuremu ari pedyo. Mushure mezvo, tinoverengera mutsauko pakati pehurefu hweaxis hwemabhokisi ekusungirira mashoma ematanho ari pedyo ane nhamba yakafanana. Kana kukosha uku kuri kushoma kana kuenzana nechikamu chehurefu chakatarwa, zvidimbu zvinogovaniswa muboka rimwechete. Zvikasadaro, tinoenzanisa mutsauko mukureba kweakisisi yenzira huru yebhokisi rinosungirira diki muchikamu chinowirirana chefuremu yakatarisana neyakatsanangurwa dhizaini pamberi uye shure kwenhamba iri mufuremu yakatarisana. Mutsara une musiyano mudiki unobva wapatsanurwa muboka rimwechete. Izvi zvinovimbisa kupatsanurwa kwakanaka pakati pematanho ari padyo, uye kuitira kuti uone kugovanisa kwakanaka kwehukama hwepedyo.
2. Spatio-temporal mamiriro-inotungamirwa kufanotaura maitiro
   Zvichienderana nechikamu, tinoshandisa uye nekuwedzera maitiro ekufungidzira anotaurwa muChikamu 3.3. Isu tinobatanidza inter-frame context mukuita, zvichireva kuti yekutanga layer yeboka rega rega, iyo inoshanda seR-layer, inogona kunge isingaite yakanakisa kufanotaura mhedzisiro. Kuti tiongorore zvizere hukama hunogona kuitika pakati pematanho ari padyo, isu tinofanirwa kuburitsa iyo yakakwana yekufanotaura maitiro.
   Chekutanga, isu tinoverenga zvakasara zvekufungidzira kune yega yega yakachekwa-layer muboka razvino kana ichishandiswa seR-layer. Nekuenzanisa zvisaririra zvekufungidzira mune zvese zviitiko, isu tinosarudza iyo R-layer ine diki diki inosara kukosha seyakanyanya kufanotaura maitiro. Kune R-layer ipfupi nzira yekuverenga, isu tinoshandisa anofamba mutengesi algorithm kuverengera ipfupi nzira yeR-layers pasi peyakanyanya kufanotaura maitiro. Uyezve, isu tinoverenga zvakasara zvekufungidzira zveboka rega rega pasi pemamodhi avo ekufanotaura. Isu tinorekodha zvakare hurefu hwekugara uye R-layer ruzivo rweboka rega rega kuti riwedzere kudzvanywa mukuzotevera kugadzirisa. Mukuita kwekutevera, isu tinoshandisa arithmetic coding zvichibva pane yakanakisa mamiriro ekusarudzwa kweruzivo rwuri pamusoro kupedzisa maitiro ese e-multi-frame point Cloud geometry compression algorithm.

Arithmetic Coding Kubva pane Context Duramazwi
Huwandu hukuru hwemamiriro ezvinhu mupoint cloud hunoremedza zvakanyanya iyo yakazara compression scheme maererano ne arithmetic coding computational kuomarara. Isu tinovandudza arithmetic coding kubva kumamodule maviri anotevera. 1) Isu tinomisikidza duramazwi rechirevo, uye tosarudza nekuvandudza kukosha kwepasirese zvinoenderana nefungidziro ye entropy, uyezve 2) tinotora encoder anochinjika kuti tiverenge zvine hungwaru miganho yekumusoro neyepasi yezvinobvira.

1. Context kuvakwa kweduramazwi
2. Isu tinogadzira duramazwi rechirevo chechinyorwa rinomiririra mutsetse utatu, unosanganisira makongiresi egore remapoinzi pachikamu chega chega chakachekwa uye nhamba inomiririra yemamiriro ayo asina chinhu. Nekudaro, isu tinosanganisa mavhokseli ari mupoindi yegore nekadiki-diki bhokisi rinosungirirwa rechikamu chimwe nechimwe nemamiriro ayo asina chinhu. Kuenzanisira kuvakwa kwemutsara wemitsara mitatu yeduramazwi rechirevo zvakajeka, tinopa tsananguro inonzwisisika mumufananidzo 2. Pamakona maviri ane mimvuri mumufananidzo 2, chete mamepu emamiriro ezvinhu pc1 uye pc2 anotariswa. Mupiro wemamiriro ezvinhu pamwe ne x-axis uye y-axis inorekodhwa kumitsara miviri QX - uye QY - zvichiteerana. Saka duramazwi remukati rine QX - uye QY -. Zvinhu zvemumutsara zvine makongiresi mamwe chete zvinosanganiswa kuita katatu, iyo inomiririrwa yakazara iyo inoverengerwa sehuwandu hwezvipo zvezvakabatanidzwa zvabatanidzwa katatu.
   Naizvozvo, mamiriro evoxel ega ega anogona kuverengerwa sehuwandu hwezvipo zvakazvimiririra zvemavoxel anogarwa muduramazwi raro. Chimiro ichi chinobatsira kuona kana voxel ichifanirwa kuwedzerwa kuduramazwi rechirevo pasina anonetesa matrix lookups, zvichikonzera kudzikiswa kwakakura mukuoma kwemakomputa uye nguva yekumhanya.
3. Probability calculation
   Kuti uverenge mukana we entropy, zvese kureba kwekutevedzana uye mamiriro ezvakaiumba ma voxels zvinofanirwa kuverengerwa. Mumodule iyi, tinogadzira encoder inogadzirika iyo inotanga inofungidzira yekumusoro uye yakaderera kuwedzeredza mukana weboka rega rega kubva muduramazwi rechirevo, uye yozoiisa encode zvinotevera. Chekutanga pane zvese, isu tinovaka bhinari muti wakavakirwa paMarkov ketani modhi. Nekufamba nekugara kwemavoxels, tinopa kukosha kwe1 uye 0 kune akagarwa uye asina chinhu mavoxels, zvichiteerana, uye kuverenga mukana zvichienderana nemuti chimiro. Kutanga kubva pamudzi node, kana voxel yakagarwa, tinonyora ruboshwe rwemwana node se 1. Kana zvisina kudaro, tinoisa mutsara wekodzero yemwana se 0 uye tinopfuurira kune imwe nhanho yekutonga nekuparadzanisa. Iyo yekuverengera fomula yekumhanya mukana wemavoxels akagarwa anogona kuwanikwa muEq. (4).

Pahurefu hwekumhanya isingasviki kana kuenzana nan, panogona kunge paine 2n yemanodhi emuti anomiririra nzvimbo dzemavoxels. Naizvozvo, mukana wevoxel chero yakagarwa inomiririrwa neyakazvimiririra yemajoini mukana wekufamba matunhu ese kutangira pamudzi uye kugumira chero isina mwana node yemuti. Kubva paEq. (4), kuita arithmetic encoding pakugara kwevoxel kutevedzana, tinoda kuwanda kwepamusoro nepasi zvingaita zvekutevedzana, sezvakaratidzwa muEq. (5).

Tichishandisa nzira iyi, tinokwanisa kushandisa zvigadziriso zvemasvomhu ekodheti kuti tigadzirise kukosha kwefungidziro yechiratidzo chega chega zvichibva pane yakagadziridzwa mukana wekufungidzira modhi uye kuwanda kwechiratidzo chega chega mukutevedzana kwemazuva ano. Izvi zvinotitendera kuverengera ekumusoro neakadzika maganhuro ekuwedzera mukana wemavoxels akagarwa uye kupedzisa encoding process.

Experiment

Tsanangudzo Yekuita

1. Dataset. Kuona mashandiro enzira yedu yatakaronga, kuyedza kwakakura kwakaitwa pamusoro pegumi nematanhatu dhatabheti yegore inogona kutorwa kubva kuRef. [16], sezvakaratidzwa mumufananidzo 40, umo maFig. 3(a)– 3(l) mifananidzo ine dense point, uye Fig. 3(m) - 3(p) zvivakwa zvine mapoinzi mashoma. Figs. 3 (a) - 3 (h) inoratidzirwa miviri yepamusoro inonongedza gore data sequences yezvigadziriso zviviri zvenzvimbo zvakatorwa kubva kuMicrosoft. Figs. 3(i)– 3(l) vanosarudzwa kubva pa3i voxelized full body point Cloud data sequences. Kusara hombe-yakakura sparse point makore muFigs. 8(k)– 3(p) inomira facade uye dhatabheti rekuvaka.
2. Evaluation metrics. Kuita kweiyo nzira yakarongwa kunoongororwa maererano nebit per point (BPP). BPP inoreva huwandu hwezvimedu zvakagarwa neruzivo rwekuronga rwakanamirwa pane iyo poindi. Iyo yakaderera kukosha, zviri nani kuita. uko Sizedig inomiririra huwandu hwezvimedu zvakagarwa neiyo coordinate ruzivo rwepoint Cloud data, uye k inoreva huwandu hwemapoinzi mune yekutanga point gore.
3. Benchmarks. Isu tinonyanya kuenzanisa nzira yedu nemamwe maitiro ekutanga, kusanganisira: PCL-PCC: octree-based compression muPCL; G-PCC (MPEG intra-coders test model) uye interEM (MPEG inter-coders test model) inotarisa imwe-frame uye yakawanda-frame point cloud compression zvichiteerana; Iyo Silhouette 3D (S3D) [41] uye Silhouette 4D (S4D) [42] inonangwa imwe furemu uye akawanda-frame point cloud compression, zvichiteerana.
   KuPCL, tinoshandisa octree point cloud compression approach muPCL-v1.8.1 yekutsikirira geometry chete. Isu tinoseta octree resolution parameters kubva pakurongeka chaiko uye voxel resolution. Kune G-PCC (TM13-v11.0), tinosarudza geometry isina kurasikirwa -isina kurasikirwa hunhu mune octree-predictive modhi, ichisiya paramita seyakagadzika. Kune interEM (tmc3v3.0), tinoshandisa mhedzisiro yekuedza pasi pekurasikirwa kwejometri uye kurasikirwa kwehunhu mamiriro sekuenzanisa[43]. Pamusoro peS3D
   uye S4D, isu tinotevera default mamiriro uye paramita.
4. Hardware. Iyo yakarongwa algorithm inoshandiswa muMatlab neC ++ uchishandisa mamwe mabasa ePCL-v1.8.1. Zvese zviedzo zvakaedzwa palaptop ine Intel Core i7- 8750 CPU @2.20 GHz ine 8 GB memory.

Mibairo yeSingle-Frame Point Cloud Compression

1. Kudzvanya mhedzisiro yemifananidzo yedense point Cloud data sequences
   Tafura 1 inoratidza mashandiro enzvimbo yedu yakatungamirwa isina kurasikirwa point Cloud geometry compression algorithms ichienzaniswa nePCL-PCC, G-PCC uye S3D nzira pamifananidzo yedense point Cloud data sequences. Zvinogona kuoneka kubva kuTable 1 kuti kune ese mapoinzi gore reakatevedzana akafanana, iyo yakarongwa nzira inowana yakaderera compression BPP kana ichienzaniswa nedzimwe nzira. Algorithm yedu inopa maavhareji ekuwana kubva -1.56% kusvika -0.02% paS3D, uye kuwana kubva -10.62% kusvika -1.45% paG-PCC. Inoratidza advan iri pachenatage, ndiko kuti, kubudirira kwekuita kwekumanikidza kweiyo yakarongwa algorithm kubva -10.62% kusvika - 1.45%; Kune PCL-PCC, iyo yakarongwa algorithm inoratidza yakapetwa kaviri kuwana pane ese akateedzana, kubva -154.43% kusvika -85.39%.
2. Kudzvanya mhedzisiro yehombe-yakakura sparse point Cloud data
   Nekuti iyo S3D haigone kushanda mune iyi kesi, isu tinongoenzanisa yedu yenzvimbo-inotungamirwa isina kurasikirwa geometry point Cloud compression algorithm nePCL-PCC uye G-PCC nzira pahombe-mwero sparse point Cloud data. Zvakare, algorithm yedu inowana kushanda kwakanyanya neG-PCC uye PCL-PCC, sezvakaratidzwa muTebhu 1. Zvigumisiro zvakaratidza kuti mavhareji eBPP anowanikwa kubva - 8.84% kusvika -4.35% anotorwa achienzaniswa neG-PCC. YePCL- PCC, yedu yakatsanangurwa algorithm inoratidza zviri pachena advantages, nemhindu dzinobvira pa -34.69% kusvika -23.94%.
3. Summary
   Kupa kuenzanisa kunonzwisisika kweiyo imwe chete furemu point cloud compression mhedzisiro, Tafura 2 inopa avhareji mhedzisiro pakati penzvimbo yedu yenzvimbo-inotungamirwa yekumanikidza nzira uye dzimwe nyika-ye-iyo-iyo-ye-benchmark nzira. Kuenzaniswa neS3D, nzira yedu yatinoda inoratidza avhareji yekuwana kubva - 0.58% kusvika - 3.43%. Kana iri G-PCC nePCL-PCC, avhareji yekuwana inowana zvishoma - 3.43% uye -95.03% zvakateerana. Ongororo yekuyedza inoburitsa kuti nzira yedu yekumanikidza yenzvimbo inodarika yazvino S3D, G-PCC uye PCL-PCC nemuganho wakakosha. Nekudaro, inogona kugutsa izvo zvisingarasikike zvekumanikidza zvinodiwa zvepoint cloud geometry yemhando dzakasiyana dzezviitiko, semuenzaniso, kugova kana kushoma kugovera, uye kushanda kwenzira yedu kunoramba kuripo.
4. Mibairo yeMulti-frame Point Cloud Compression
   Isu tinoongorora yedu yakatsanangurwa spatial-yenguva mamiriro-inotungamirwa point Cloud geometry compression algorithm inopesana nearipo compression algorithms senge S4D, PCL-PCC, G-PCC uye interEM. Mifananidzo chete yedense point Cloud data sequence ndiyo inoshandiswa mukuyedza uku. Zvigumisiro zvinoratidzwa mu.

Tafura 1. BPP kuenzanisa yenzvimbo yedu yepasi-inotungamirwa compression algorithm uye nzira dzekutanga.

Tafura 2. BPP kuenzanisa nemamiriro-ye-the-art algorithms pane imwe-frame point cloud data.

Tafura 3. Sezvatinoona, mushure mekugadzirisa mune yekufembera modhi uye arithmetic encoder, iyo yakarongwa algorithm inoratidza hukuru pane ese bvunzo kutevedzana. Kunyanya, kana ichienzaniswa neinterEM neG-PCC, iyo algorithm yakarongwa inoratidza budiriro huru kubva -51.94% kusvika -17.13% uye -46.62% kusvika -5.7%, zvichiteerana. Kuenzaniswa neS4D, iyo algorithm yakatsanangurwa inoratidza kuvandudzwa kwakasimba kubva -12.18% kusvika -0.33%. Kana iri PCL-PCC, yedu yakarongwa algorithm yakada kudzika nepakati pane ese ebvunzo akateedzana.
Uyezve, tinopfupikisa mibairo yekutsikirira uye zvakawanikwa zvenzira yakarongwa pamufananidzo wedense point Cloud data sequences, dzakanyorwa muTebhurari 4. Paavhareji, inopa zvakawanikwa pakati pe -11.5% ne -2.59% zvichienzaniswa nenzvimbo inotungamirwa yegore. geometry compression algorithm yakatsanangurwa kare. Uyezve, inoratidza avhareji yeavhareji yekuwana - 19% kana ichienzaniswa neG-PCC uye yakawana avhareji yekodhi yekodhi ye -24.55% zvichienzaniswa neinterEM. Pamusoro pezvo, zvichienzaniswa neS3D neS4D, inowana zvinopfuura -6.11% uye -3.64% paavhareji zvichiteerana. Ongororo yezvese yekuyedza inoratidza kuti spatiotemporal inotungamirwa-inotungamirwa point Cloud compression nzira inogona kushandisa zvizere zvese zvepakati uye zvenguva kuwirirana kwezvikamu zviri pedyo mukati me intra-frames uye inter-frames. Isu tinonatsurudza zvakare kusarudzwa kwemamiriro epasirese uye mukana wemodhi yearithmetic encoder kuti tiwane yakaderera. Nzira yakarongwa inodarika mashandiro echizvino-zvino-e-algorithms, kuitira kuti isangane nezvinodiwa zvepoint Cloud geometry isingarasikike compression mumultimedia application scenarios senge dynamic portraits.

Tafura 3. Bit-per-point kuenzanisa yedu spatio-temporal mamiriro-inotungamirwa compression algorithm uye yekutanga nzira.

Tafura 4. Bit-per-point kuenzanisa neye-ye-the-art algorithms pane akawanda-frame point Cloud data.

Ablation Study
Isu tinoita zvidzidzo zveablation pakufungidzira encoding pamusoro pe8i voxelized yakazara-muviri point Cloud data sequences kuratidza kushanda kwechikamu. Zvinogona kuonekwa kubva paTable 5 kuti kuvandudzwa kunoratidza kuwana kwakagadzikana kwe -70% pane multiframe point cloud compression uye - 60% pane imwe-frame point cloud compression inopesana ne-non-partition predictive coding.
Tevere, tinoita ongororo yeablation pa arithmetic coding kuratidza kushanda kwechirevo chechinyorwa. Sezvinoratidzwa muTable 6, kuvandudzwa kwakasimba kwe −33% pakutsikirirwa kwefuremu yakawanda-yakasiyana-siyana uye iyo -41% pa-single-frame point Cloud compression inopesana nemasvomhu pasina duramazwi rinoonekwa nenzira yedu.

Nguva Yekushandisa
Isu tinoyedza mashandisiro enguva kuti tiongorore kuoma kwealgorithm uye kuenzanisa nzira dzakarongwa nevamwe. Iyo algorithm yakaoma inoongororwa nema encoder uye decoder yakazvimiririra, yakanyorwa muTable 7. Sezvatinoona, G-PCC, interEM uye PCL-PCC inogona kuwana encoding nguva isingasviki 10 s uye decoding nguva isingasviki 5 s portrait-dense point Cloud data. Ivo zvakare vanoita zvakanaka mune yakakura-yakakura sparse point Cloud data kana ichienzaniswa nevamwe. Yedu yakarongwa maalgorithms anotora anenge 60 s uye 15 s encode uye decode mifananidzo inoteedzana, zvakanyanya pane facade uye yekuvaka point Cloud data. Pane kutengeserana-kusiyana pakati pe bitrates uye compression kumhanya. Kuenzaniswa neS3D uye S4D, izvo zvinotora mazana emasekonzi kuti encode, yedu inopedza nguva nzira inogona kuratidza hukuru.
Muchidimbu, mashandisirwo enguva yemaitiro edu akarongwa ari pakati nepakati pakati peese akaenzaniswa algorithms asi zvichiri kukosha kuti iwedzere kuvandudzwa.

Mhedziso

Mubepa rino, isu tinokurudzira spatiotemporal inotungamirwa nzira yekusarasikirwa point Cloud geometry compres⁃ sion. Isu tinoona yakachekwa point Cloud yeunit ukobvu seyekupinza unit uye totora iyo geometry predictive coding mode zvichibva pane anofamba mutengesi algorithm, iyo inoshanda kune zvese zviri zviviri intra-frame uye inter-frame. Zvakare, isu tinoshandisa zvizere ruzivo rwepasirese mamiriro uye adaptive arithmetic encoder zvichibva pane mamiriro-nekukurumidza kugadzirisa kuti tiwane kusarasikirwa kwekumanikidza uye decompression mhedzisiro yemakore point. Mibairo yekuedza inoratidza kushanda kwenzira dzedu uye hukuru hwadzo pane zvidzidzo zvekare. Zvebasa remangwana, tinoronga kuenderera mberi nekudzidza kuomarara kwese kwegorgorithm, nekudzikisira kuomarara kwealgorithm kuti tiwane yakakwirira-kumhanya yekumanikidza mwero uye yakaderera bit mwero wekumanikidza. Iyo yakaderera bit mwero uye chaiyo-nguva / yakaderera-kunonoka inotsigirwa nzira inodiwa zvakanyanya mumhando dzakasiyana dzezviitiko.

References

1. MI XX, YANG BS, DONG Z, nevamwe. Otomatiki 3D mugwagwa muganho kutorwa uye vectorization uchishandisa MLS point makore [J]. IEEE kutengeserana pane akangwara ekufambisa masisitimu, 2022, 23 (6): 5287 - 5297. DOI: 10.1109/ TITS.2021.3052882
2. DONG Z, LIANG FX, YANG BS, nevamwe. Kunyoreswa kwehombe-yepasirese laser scanner point makore: a review uye benchmark [J]. ISPRS magazini yephotogrammetry uye kure sensing, 2020, 163: 327– 342. DOI: 10.1016/j.isprsjprs.2020.03.013
3. GRAZIOSI D, NAKAGAMI O, KUMA S, et al. An overview yeinoenderera mberi point cloud compression standardization zviitiko: vhidhiyo-based (V-PCC) uye geometry-based (G-PCC) [J]. APSIPA kutengeserana pachiratidzo uye kugadzirisa ruzivo, 2020, 9: e13
4. DE QUEIROZ RL, CHOU P A. Dzvinyiriro ye 3D point makore uchishandisa dunhu-adaptive hierarchical transform [J]. IEEE kutengeserana pakugadzirisa mufananidzo, 2016, 25 (8): 3947–3956. DOI: 10.1109/TIP.2016.2575005
5. BLETTERER A, PAYAN F, ANTONINI M, et al. Point Cloud compression uchishandisa kudzika mepu [J]. Mifananidzo yemagetsi, 2016, 2016(21):1–6
6. MEKURIA R, BLOM K, CESAR P. Dhizaini, kuita, uye kuongororwa kweiyo point Cloud codec yetele-immersive vhidhiyo [J]. IEEE kutengeserana pamaseketi uye masisitimu evhidhiyo tekinoroji, 2017, 27(4): 828 - 842. DOI: 10.1109/ TCSVT.2016.2543039
7. DE QUEIROZ RL, CHOU P A. Kutsikirira-kunotsiviwa kwemafuremu anoshanduka-shanduka [J]. IEEE kutengeserana pakugadzirisa mufananidzo, 2017, 26 (8): 3886–3895. DOI: 10.1109/TIP.2017.2707807
8. CAO C, PREDA M, ZAHARIA T. 3D point cloud compression: ongororo [C]// The 24th International Conference on 3D Web Vadivelu Comedy Technology. ACM, 2019: 1–9. DOI: 10.1145/3329714.3338130
9. GRAZIOSI D, NAKAGAMI O, KUMA S, et al. An overview yeinoenderera mberi point cloud compression standardization zviitiko: vhidhiyo-based (V-PCC) uye geometry-based (G-PCC) [J]. APSIPA kutengeserana pachiratidzo uye kugadzirisa ruzivo, 2020, 9(1): e13. DOI: 10.1017/atsip.2020.12
10. HUANG Y, PENG JL, KUO CJ, nevamwe. Octree-based progressive geometry coding of point clouds [C]//The 3rd Eurographics/IEEE VGTC Musangano paPoint-Based Graphics. IEEE, 2016: 103–110
11. FAN YX, HUANG Y, PENG J L. Kutsikirira kwegore kunobva pane hierarchical point clustering [C]//Asia-Pacific Signal and Information Processing Association Annual Summit and Conference. IEEE, 2014: 1 - 7. DOI: 10.1109/APSIPA.2013.6694334
12. DRICOT A, ASCENSO J. Adaptive multi-level triangle soup for geometry-based point cloud coding [C]//The 21st International Workshop on Multimedia Signal Processing (MMSP). IEEE, 2019: 1 - 6. DOI: 10.1109/ MMSP.2019.8901791
13. HE C, RAN LQ, WANG L, nevamwe. Point set surface compression zvichibva pakuongororwa kwechimiro chechimiro [J]. Multimedia maturusi nemashandisirwo, 2017, 76(20): 20545–20565. DOI: 10.1007/s11042-016-3991-0
14. IMDAD U, ASIF M, AHMAD M, et al. Three-dimensional point Cloud compression uye decompression uchishandisa polynomials yedhigirii yekutanga [J]. Symmetry, 2019, 11(2): 209. DOI: 10.3390/sym11020209
15. SUN XB, MA H, SUN YX, et al. A novel point Cloud compression algorithm yakavakirwa pakubatana [J]. IEEE marobhoti uye otomatiki mavara, 2019, 4 (2): 2132-2139. DOI: 10.1109/LRA.2019.2900747
16. DE OLIVEIRA RENTE P, BRITES C, ASCENSO J, et al. Graph-based static 3D point makore geometry coding [J]. IEEE kutengeserana pamultimedia, 2019, 21(2): 284-299. DOI: 10.1109/TMM.2018.2859591
17. ISO. Geometry-based point cloud compression (G-PCC): ISO/IEC 23090-9 [S]. 2021
18. DRICOT A, ASCENSO J. Hybrid octree-ndege point cloud geometry coding [C]//The 27th European Signal Processing Conference (EUSIPCO). IEEE, 2019: 1–5
19. ZHANG X, GAO W, LIU S. Implicit geometry partition for point cloud compression [C]// Proceedings of 2020 Data Compression Conference (DCC). IEEE, 2020: 73–82. DOI: 10.1109/DCC47342.2020.00015
20. QUACH M, VALENZISE G, DUFAUX F. Kudzidza convolutional inoshandura yerasika point cloud geometry compression [C]//The 2019 IEEE International Conference on Image Processing (ICIP). IEEE, 2019: 4320– 4324. DOI: 10.1109/ICIP.2019.8803413
21. HUANG TX, LIU Y. 3D point cloud geometry compression pakudzidza kwakadzama [C]//The 27th ACM International Conference on Multimedia. ACM, 2019: 890–898. DOI: 10.1145/3343031.3351061
22. GUARDA AFR, RODRIGUES NMM, PEREIRA F. Point cloud coding: Kutora yakadzama yekudzidza-based approach [C]//Picture Coding Symposium (PCS). IEEE, 2020: 1–5. DOI: 10.1109/PCS48520.2019.8954537
23. WANG JQ, ZHU H, MA Z, nevamwe. Yakadzidza point cloud geometry compression [EB/OL]. [2023-09-01]. https://arxiv.org/abs/1909.12037.pdf
24. AINALA K, MEKURIA RN, KHATHARIYA B, et al. Iyo yakagadziridzwa yekuwedzera⁃ ment layer yeoctree-based point cloud compression ine ndege projection app⁃ proximation [C]//SPIE Optical Engineering+Applications. SPIE, 2016: 223– 231. DOI: 10.1117/12.2237753
25. SCHWARZ S, HANNUKSELA MM, FAKOR-SEVOM V, nevamwe. 2D vhidhiyo kukodha yevhoriyamu data data [C]//Mufananidzo Coding Symposium (PCS). IEEE, 2018: 61–65. DOI: 10.1109/PCS.2018.8456265
26. FAKOR SEVOM V, SCHWARZ S, GABBOUJ M. Geometry-inotungamirwa 3D kududzira data yekufungidzira-based dynamic point Cloud coding [C]//The 7th European Workshop paVisual Information Processing (EUVIP). IEEE, 2019: 1–6. DOI: 0.1109/EUVIP.2018.8611760
27. KATHARIYA B, LI L, LI Z, et al. Kusarasikirwa kwesimba kwegore geometry compression ine inter-muripo uye anofamba mutengesi kufanotaura [C]// Data Compression Conference. IEEE, 2018: 414. DOI: 10.1109/ DCC.2018.00067
28. ISO. Visual volumetric video-based coding (V3C) uye vhidhiyo-based point Cloud compression: ISO/IEC 23090-5 [S]. 2021
29. PARK J, LEE J, PARK S, nevamwe. Projection-based occupancy mepu coding ye 3D point Cloud compression [J]. IEIE transaction pane smart process & computing, 2020, 9 (4): 293-297. DOI: 10.5573/ieiespc.2020.9.4.293
30. COSTA A, DRICOT A, BRITES C, et al. Yakavandudzwa chigamba packing yeMPEG V-PCC chiyero [C]//IEEE 21st International Workshop paMultimedia Signal Processing (MMSP). IEEE, 2019: 1 - 6. DOI: 10.1109/ MMSP.2019.8901690
31. KAMMERL J, BLODOW N, RUSU RB, et al. Chaiyo-nguva yekutsikirira kweiyo point cloud streams [C]// Proceedings of 2012 IEEE International Conference on Robotics and Automation. IEEE, 2012: 778 - 785. DOI: 10.1109/ ICRA.2012.6224647
32. PCL. Point Cloud library. [EB/OL]. [2023-09-01]. http://pointclouds.org/
33. THANOU D, CHOU PA, FROSSARD P. Graph-based compression ye dynamic 3D point cloud sequences [J]. IEEE kutengeserana pakugadzirisa mufananidzo, 2016, 25 (4): 1765-1778. DOI: 10.1109/TIP.2016.2529506
34. LI L, LI Z, ZAKHARCHENKO V, et al. Yepamberi 3D mafambiro ekufanotaura kwevhidhiyo-based point cloud attributes compression [C]//Data Compression Conference (DCC). IEEE, 2019: 498–507. DOI: 10.1109/DCC.2019.00058
    ZHAO LL, MA KK, LIN XH, nevamwe. Real-time LiDAR point cloud compression uchishandisa Bi-directional kufanotaura uye range-adaptive floating-point coding [J]. IEEE kutengeserana pakutepfenyura, 2022, 68(3): 620 - 635. DOI: 10.1109/TBC.2022.3162406
35. LIN JP, LIU D, LI HQ, nevamwe. M-LVC: Mafuremu akawanda kufanotaura kweakadzidza vhidhiyo compression [C]//IEEE/CVF Musangano paComputer Vision uye Kuzivikanwa Kwepateni. IEEE, 2020: 3543 - 3551. DOI: 10.1109/ CVPR42600.2020.00360
36. YANG R, MENTZER F, VAN GOOL L, et al. Kudzidzira kudzvanya kwevhidhiyo nemhando yepamusoro uye inodzokororwa kuwedzera [C]//IEEE/CVF Musangano paComputer Vision uye Kuzivikanwa Kwepateni. IEEE, 2020: 6627–6636.DOI: 10.1109/CVPR42600.2020.00666
37. KAYA EC, TABUS I. Kutsikirira kusingarasikike kwepoint cloud sequences uchishandisa kutevedzana kwakagadziridzwa CNN modhi [J]. IEEE kuwana, 2022, 10: 83678 -83691. DOI: 10.1109/ACCESS.2022.3197295
38. DING S, MANNAN MA, POO A N. Yakatarisana nebhokisi rinosungirirwa uye octree-based global interference diagnosis mu 5-axis machining yemahara-fomu nzvimbo [J]. Dhizaini yekombuta, 2004, 36(13): 1281-1294
39. ALEXIOU E, VIOLA I, BORGES TM, et al. Chidzidzo chakazara cheyero-kukanganisa kuita muMPEG point Cloud compression [J]. APSIPA kutengeserana pachiratidzo uye kugadzirisa ruzivo, 2019, 8: e27. doi:10.1017/ ATSIP.2019.20
40. PEIXOTO E. Intra-frame compression yepoint cloud geometry uchishandisa dyadic decomposition [J]. IEEE Signal Processing Letters, 2020, 27: 246–250. DOI: 10.1109/LSP.2020.2965322
41. RAMALHO E, PEIXOTO E, MEDEIROS E. Silhouette 4D ine mamiriro ekusarudza: isina kurasikirwa geometry kudzvanywa kweiyo dynamic point makore [J]. IEEE Signal Processing Letters, 2021, 28: 1660 - 1664. DOI: 10.1109/ lsp.2021.3102525
42. ISO. Mamiriro ekuedzwa akajairika eG-PCC gwaro N00106: ISO/IEC JTC 1/ SC 29/WG 7 MPEG [S]. 2021

Nhoroondo dzehupenyu

• ZHANG Huiran akamugamuchira madhigirii eBE neME muChikoro cheGeodesy neGeomatics uye State Key Laboratory yeRuzivo Injiniya muKuongorora Mepu uye Remote Sensing, zvese kubva kuWuhan University, China muna 2020 na2023, zvichiteerana. Parizvino ndiye muongorori weGuangzhou Urban Planning uye Design Survey Research Institute, China. Zvido zvake zvekutsvagisa zvinosanganisira point Cloud data processing uye compression. Akatora chikamu mumapurojekiti akati wandei ane chekuita nemunda wekunzwa kure uye akaburitsa bepa rimwe muGeomatics uye Information Science yeWuhan University.
• DONG Zhen (dongzhenwhu@whu.edu.cn) akagamuchira madhigirii ake eBE nePhD mukunzwa kure uye mafotogiramu kubva kuWuhan University, China muna 2011 na2018, zvichiteerana. Ndiye purofesa ane State Key Laboratory yeRuzivo Injiniya muKuongorora, Mepu uye Remote Sensing (LIESMARS), Wuhan University. Zvido zvake zvekutsvagisa zvinosanganisira kuvakazve 3D, kunzwisisa kwechiitiko, point cloud processing pamwe nekushandisa kwavo muhungwaru hwekufambisa system, maguta mapatya edhijitari, budiriro yemadhorobha uye marobhoti. Akagamuchira kukudzwa kwe10 kubva kumakwikwi akasiyana-siyana enyika uye epasi rose uye akadhinda akatenderedza mapepa makumi matanhatu mumagazini akasiyana siyana nemisangano.
  WANG Mingsheng akawana dhigirii rake reBE muKoreji yeComputer Science neTekinoroji kubva kuJilin University, China muna 2001, uye dhigirii reME muChikoro cheComputer Science and Engineering kubva kuSouth China University of Technology, China muna 2004. Parizvino ndiinjiniya mukuru weGuangzhou Urban Planning. & Dhizaini Survey Research Institute, China. Zvido zvake zvekutsvagisa zvinosanganisira komputa application uye software, physiography, uye kuongorora. Akagamuchira kukudzwa kwe20 kubva kumakwikwi akasiyana-siyana enyika uye akadhinda akatenderedza mapepa makumi mashanu mumagazini akasiyana siyana nemisangano.

DOI: 10.12142/ZTECOM.202304003
https://kns.cnki.net/kcms/detail/34.1294.TN.20231108.1004.002.html, yakabudiswa paIndaneti November 8, 2023
Manyoro akagamuchirwa: 2023-09-11

Zvinyorwa / Zvishandiso

ZTE Inotungamirwa Algorithm yeLossless Point Cloud Geometry Compression [pdf] Bhuku reMushandisi Yakatungamirirwa Algorithm yeKurasika Poindi Cloud Geometry Compression, Inotungamirwa, Algorithm yeLossless Point Cloud Geometry Compression, Lossless Point Cloud Geometry Compression, Point Cloud Geometry Compression, Geometry Compression

References

• User Manual