In order to improve the prediction speed of HEVC encoder, this paper proposed an algorithm of PU size decision based on information entropy for HEVC (High Efficiency Video Coding) intra prediction. This algorithm mainly utilizes the characteristic of image entropy producing different values with the richness of the image detail to decide whether the PUs need segmentation. Firstly, we calculate the information entropy of every LCU, then we can obtain the optimal partition size from the PU size of 64×64 to 4×4, so that the optimal coding depth is acquired, we can make the intra angular prediction in the optimal coding depth. The experimental results show that we can avoid the unnecessary RD cost calculating process in the traversal search mode, and the coding time was reduced by about 22.6% in the promise of coding performance essentially unchanged. This algorithm simplified the unnecessary rate-distortion cost procedure in traversal search mode, shortened the intra prediction coding time at high image quality level.