Improving memory bandwidth efficiency of multimedia system and achieving high-definition real-time video encoding are the recent design trends in power-aware portable/vehicle/surveillance video applications, but the focus was only on low-cost, low-resolution and low-power design in the past. Furthermore, it takes a lot of hardware for high-definition video encoding associated with high external memory bandwidth and complexity to speed up to meet the requirement of real-time video encoding. As a result, developing a SOC video system that takes into account both low-cost/low-power design and high-definition real-time video encoding for power-aware video applications is a daunting challenge. This dissertation presents a low-cost bandwidth-efficient SOC architecture of H.264 video encoder with dynamic quality adjustability for power-aware video applications. We implement optimization methods in algorithm, H.264 video encoder architecture and video SOC system, respectively. First, in algorithm optimization, we propose dynamic quality-adjustable algorithms for motion estimation and intra prediction that can dynamically configure the encoding modes with the design trade-off between power consumption and video quality for various video encoding applications. Because of the setting, computational complexity can be reduced by as high as 90%. We also propose two-stage fast MB-skip algorithm without recalculating SATD value of skip cost. The algorithm can be applied to static surveillance video applications, and it has 60% of bit-rate saving in the same PSNR value as compared to those without. Second, in H.264 video encoder architecture optimization, to simplify encoding flow and eliminate data dependence between memory data accessing and kernel core processing, we adopt six pipeline stages with the order of data pre-loading, integer motion estimation, fractional motion estimation, intra prediction, entropy coding/deblocking filtering and extended output frame buffering. In addition, we design a pre-load scheduler and an extended output frame buffer module for accessing source YUV, reference data and reconstructed data to achieve high efficiency data access with a MB-based scan order. According to the abovementioned optimization methods for H.264 video encoder architecture, we reduce external memory bandwidth by 70% by decreasing the number of R/W access from external memory, and enable H.264 video encoder to operate within a reasonable range in external memory bandwidth when targeting at high-definition video applications. In SOC video system optimization, we develop AXI-like system bus and dual direct memory access (DMA) architecture to decrease the latency of external memory accessing by using a bank-interleaving technique of DRAM, achieving 100% improvement for external memory bandwidth efficiency. With 70% of reduction of external memory bandwidth and 90% of computational complexity, the proposed H.264 video encoder SOC design can satisfy real-time performance requirement of HD (1280×720@30fps) and HD1080 (1920x1080@30fps) when operated at 90MHz and 160MHz at the cost of 485K gates and 15.2K bytes local memory according to TSMC 0.13um CMOS technology.