With　the　rapid　development　of　3D　point　clouds,　it　can　better　represent　the　surface　of　objects　in　real　life.　However,　the　data　volume　of　point　cloud　is　huge,　which　requires　an　effective　compression　algorithm　to　compress　the　point　cloud.　In　this　paper,　we　propose　a　point　cloud　attribute　compression　method　based　on　bidirectional　prediction　and　color　weighted　graph　transformation　(CWGT).　First,　the　k-d　tree　is　designed　for　the　partition　processing　of　the　point　clouds.　This　hierarchical　representation　provides　roughly　the　same　number　of　levels　for　each　leaf　node.　Second,　we　introduce　a　block-based　bidirectional　intra　prediction　scheme,　which　proposes　a　reasonable　prediction　model　to　eliminate　spatial　redundancy.　Finally,　in　order　to　adapt　to　the　irregularity　of　the　point　cloud,　we　apply　graph　transformation　to　deal　with　the　prediction　residual　of　the　transform　block,　where　the　edge　weight　of　the　graph　is　weighted　by　the　color　attribute　and　the　geometric　space　distance.　The　experimental　results　show　that　the　average　bit　rate　of　this　scheme　is　reduced　by　about　24.58%　on　the　upper　body　dataset.　In　terms　of　visual　quality,　it　can　effectively　improve　the　fineness　of　point　cloud　model　compression　reconstruction.　©　2022　IEEE.