Abstract: To address the bad effect of intra-coding Rate Control (RC) of H.264, a novel image complexity adaptive I-frame RC algorithm is proposed. First, the gradient of luma pixel is detected in I-frame with Sobel operator and the edge direction histogram is established for each 44 block, hereby the most probable intra prediction mode and corresponding reconstructed block are got. Finally, the residual picture which is close to the actual coding residual is obtained. The mean absolute value of residual is used to represent I-frame coding complexity, then an empirical Rate-Quantization (R-Q) model is proposed, and the optimal intra-QP is accurately determined for each GOP according to allocated target bits by considering simultaneously buffer status and sequence characteristic. Experimental results show that the proposed scheme obtains more accurate I-frame output bit-rate and more steady video quality. Buffer overflow and frame skip are effectively prevented, and sequence PSNR fluctuation has reduced by 60%.