We present several efficient adaptive redundant picture coding methods for error resilient video coding. In our previous work, redundant picture coding in combining with reference picture selection, reference picture list reordering and hierarchical redundant picture allocation was proposed. This paper investigates how to allocate redundant pictures more efficiently according to the content characteristics of the primary pictures. Simulation results show that the adaptive redundant picture coding methods can achieve average PSNR improvements around 2 to 4 dB compared to the loss-aware rate distortion optimized (LA-RDO) intra macroblock refresh implemented in H.264/AVC Joint Model (JM). This paper also asserts that the methods do not introduce any additional end-to-end delay, therefore suit for low-delay applications such as video telephony and video conferencing, which demand better error resilience than other applications.