With too high computational complexity brought by intra prediction in high efficiency video coding (HEVC) standard tak en into account,a low complexity coding algorithm guided by largest coding unit (LCU) similarity is proposed for HEVC in this paper.In the first place,similarit y among adjacent LCUs is analyzed in detail,which is subsequently quantified according to the detailed analysis.The quantized LCU si milarity is employed to predict the range of depth of LCUs which are currently coded.Specifically,Bayesian principle is used to guide the prediction of LCU depth,where an online update process is introduced to obtain the prior probabil ity necessary for Bayesian decisions.To further reduce encoding time,Lagrangian cost thresholds for vario us coding unit sizes are predicted adaptively by virtue of optimal Lagrangian costs of surrounding coding units.Th e thresholds are responsible whether recursive splitting of coding units of a ce rtain size can be terminated early or not.Experimental results show that compared with the standard HEVC coding scheme,the encoding time is red uced by about 39.55% on average with negligible increase of Bjontegaard delta bit rate (BDBR) (less than 1%) and decrease of Bjontegaard delat peak signal-to-noise ratio (BD-PSNR) (0.03dB),demonstrating the efectiveness of the LCU similarity based alg orithm.