The　Viterbi　algorithm　(VA)　is　a　kind　of　efficient　maximum　likelihood　(ML)　decoding　method　for　finite　state　signals　observed　in　noise　environments.　But　the　decoding　complexity　grows　exponentially　while　the　memory　size　increases　and　it　doesn　＇t　work　well　when　dealing　with　burst　errors.　Constrained　Viterbi　algorithm　(C-VA)　was　proposed　in　a　novel　product　code　scheme　[1]　to　overcome　the　drawback.　The　main　idea　of　constrained　Viterbi　algorithm　is　to　make　full　use　of　some　bits　which　are　known　to　be　decoded　correctly　first.　It　has　been　proved　that　once　the　reliable　constraint　condition　can　be　given,　the　C-VA　has　great　advantages　for　image　transmission.　In　this　paper,　by　adding　constraint　condition　directly　and　simply　for　C-VA　in　the　encoder,　a　new　scheme　is　proposed　for　error　resilient　transmission　of　set　partitioning　in　hierarchical　trees　(SPIHT)　coded　images.　Experiments　show　that　it　is　useful　when　the　channel　condition　is　extremely　inferior.　©　2006　IEEE.