This　paper　proposes　a　method　based　on　digital　image　processing　for　measuring　drivers＇　perception-reaction　time　(PRT)　to　the　green　phase.　The　detection　of　the　onset　of　a　green　signal　was　based　on　the　RGB　color　model.　The　detection　of　the　start　of　the　vehicle　was　based　on　the　frame　difference.　The　driver＇s　perception-reaction　time　was　equal　to　the　difference　between　the　frame　of　the　onset　of　the　green　signal　and　the　frame　of　the　start　of　the　vehicle　divided　by　the　video　frame　rate.　Drivers＇　PRT　with　and　without　a　countdown　timer　were　comparatively　analyzed.　First,　the　means　and　variances　of　drivers＇　PRT　with　a　timer　and　without　a　timer　were　compared.　Second,　Normal,　Lognormal,　Gamma　and　Weibull　distributions　were　used　to　fit　the　PRT　data.　Third,　the　fuzzy　c-means　clustering　was　utilized　for　PRT　classification　and　comparison.　The　results　show　that　the　method　for　measuring　drivers＇　PRT　based　on　digital　image　processing　was　effective.　The　drivers＇　PRT　was　decreased　from　2.12　s　to　1.48　s　with　countdown　signals.　Weibull　distribution　appeared　to　best　fit　the　PRT　with　No-Timer　and　Lognormal　distribution　appeared　to　best　fit　the　PRT　with　Timer.　The　PRT　data　were　classified　into　three　groups:　＂Fast＂,　＂Moderate＂　and　＂Slow＂.　Clustering　centers　without　Timer　were　1.16　s,　2.47　s　and　3.98　s　respectively.　Clustering　centers　with　Timer　were　0.95　s,　1.64　s　and　2.70　s　respectively.　The　clustering　centers　can　be　regarded　as　reference　values　of　drivers＇　starting　response　for　microscopic　traffic　simulation　software.　(C)　2013　Elsevier　Ltd.　All　rights　reserved.