Scalable video coding (SVC) has been standardized to extend the capabilities of the H.264 advanced video coding (AVC). The SVC can compress several video sequences of various resolutions as a single bit-stream. In the SVC enhancement layer, for Joint Scalable Video Model (JSVM) software implementation, an exhaustive mode decision process based on the base layer mode predictions is performed to obtain the best mode for each macroblock (MB). This technique may achieve a higher coding efficiency; however, it induces a significant computational complexity in the encoding engine. In order to speedup the SVC encoder, a fast mode decision algorithm was proposed in this paper. In other words, our aim was to decrease the number of candidate modes to reduce the computational complexity and maintain the same level of coding efficiency, this approach used the spatial and temporal correlation between MB situated at the enhancement layer and its co-located MB at the base layer. Our statistical analyses were made using several HD sequences with different motion characteristics. Experimental results show a significant improvement in terms of time encoding which is a major constraint for any real-time implementation. However, this gain is accompanied with an acceptable loss in video quality and a tolerable bit rate increase for most media supports. In fact, our proposed algorithm permits a major improvement that can reach up to 64.9 % in terms of computational effort. This gain will induce an average loss yield to 10.5 or 13.87 % that is comparable to the 13.12 % of the He Li's algorithm with an acceptable loss in terms of subjective video quality.