Terahertz　(THz)　radiation　lies　between　the　millimeter　and　infrared　region　of　the　electromagnetic　spectrum,　which　is　typically　defined　as　the　frequency　range　of　0.1-10　THz　and　the　corresponding　wavelength　ranges　from　30　mu　m　to　3　mm.　Terahertz　radiation　due　to　wide　spectrum,　high　penetration,　low　energy,　and　other　important　features,　has　been　a　valuable　tool　for　imaging　and　non-destructive　testing　on　a　submillimeter　scale.　Continuous-wave　(CW)　terahertz　ptychography　is　a　type　of　phase-contrast　technique　with　advantages　of　simple　set-up　and　large　field-of-view.　It　retrieves　the　complex-valued　transmission　function　of　the　specimen　and　the　probe　function　at　the　same　time.　The　extended　ptychographic　iterative　engine　(ePIE)　algorithm　is　used　as　the　reconstruction　algorithm　in　the　field　of　ptychography,　because　it　is　relatively　simple,　and　can　use　computer　memory　efficiently.　However,　the　problem　of　algorithm　convergence　delay　makes　us　unable　to　acquire　the　reconstruction　result　very　quickly.　Since　the　ptychography　is　a　problem　of　retrieving　phase　information,　physical　constraints　affect　the　convergence　speed　of　the　algorithm　strongly.　In　this　paper,　we　propose　a　dual-plane　ePIE　(dp-ePIE)　algorithm　for　CW　THz　ptychography.　By　moving　detector　along　the　axis　and　capturing　diffraction　patterns　of　one　zone　of　an　object　at　two　recording　planes,　then,　two　sets　of　patterns　used　as　the　constraints　simultaneously　can　increase　the　diversity　of　experimental　parameter.　Hence,　the　convergence　rate　can　be　improved.　The　simulation　results　suggest　better　reconstruction　fidelity　with　a　faster　convergence　rate　by　the　dp-ePIE　algorithm.　The　dual-plane　terahertz　ptychography　experimental　setup　is　built　based　on　2.52　THz　optically　pumped　laser　and　Pyrocam-III　pyroelectric　array　detector.　Compared　with　other　methods　to　increase　the　diversity　of　measurement,　the　setup　of　dual-plane　ptychography　can　be　compact　and　simple,　thus　reducing　the　terahertz　wave　transmission　loss.　A　polypropylene　sample　is　adopted　and　it　is　approximated　as　a　pure　phase　object.　No-reference　structural　sharpness　(NRSS)　is　utilized　as　a　quantitative　evaluation　index.　It　takes　45.086　s　to　achieve　NRSS　value　of　0.9831　by　using　the　dp-ePIE　algorithm　in　10　iterations,　while　the　NRSS　value　and　calculation　time　for　e-PIE　algorithm　are　0.9531　and　57.117　s　(20　loops),　respectively.　The　experimental　results　show　that　the　dp-ePIE　algorithm　can　obtain　high-quality　amplitude　and　phase　distribution　with　less　iterations　than　the　traditional　ePIE　algorithm.