A　novel　visually　meaningful　image　encryption　algorithm　is　proposed　based　on　adaptive　2D　compressive　sensing　and　chaotic　system.　The　plain　image　is　first　compressed　and　encrypted　simultaneously　by　adaptive　2D　compressive　sensing　to　obtain　the　pre-encrypted　compressed　image.　In　this　process,　3D　cat　map　is　used　to　generate　the　measurement　matrix　and　the　scrambling　sequence.　Then,　the　pre-encrypted　compressed　image　is　embedded　into　the　host　image　by　dynamic　LSB　method　based　on　2(K)　correction　so　as　to　get　cipher　images　with　higher　visual　quality.　A　four-dimensional　discrete　chaotic　system　is　used　for　region　scrambling　in　the　embedding　process　in　order　to　further　improve　the　security　of　the　algorithm.　In　the　simulation　tests,　the　plain　image　with　the　maximum　size　2048　x　2048　can　be　compressed　and　embedded　into　a　host　image　of　size　512　x　512.　The　embedding　ratio　is　better　than　the　existing　algorithms.　Most　importantly,　our　algorithm　is　tens　or　even　hundreds　of　times　more　efficient　than　other　algorithms.