Air　pollution　is　a　crucial　environmental　problem,　especially　the　fine　particulate　matter　(PM2.5)　which　has　become　one　of　the　focal　points.　PM2.5　is　a　complex　pollutant　which　can　intrude　the　lungs　and　threaten　people＇s　health　during　the　whole　lives.　In　order　to　enable　people　to　know　the　PM2.5　index　of　their　surroundings　at　any　time,　an　image-based　PM2.5　predictor　with　saliency　detection　(IPPS)　is　proposed.　The　proposed　predictor　first　obtains　the　non-salient　regions　based　on　saliency　detection　technologies.　Then,　the　authors　extract　two　features　of　the　entropy　and　intensity　values　of　non-salient　image　saturation　map.　Finally,　they　multiply　these　two　features　into　the　approximation　of　PM2.5　concentration.　Experiments　show　that　the　proposed　IPPS　is　superior　in　accuracy　and　efficiency.