This paper presents an effective interpolation restoration algorithm for color images in Bayer format to reduce the computational complexity and improve its real time property. Based on the characteristics of human visual system in the interpolation process, the more accurate brightness information and edge information of the image can be obtained. Owing to edge information, more accurate values of|R, G, B can be recovered. The final solution of the algorithm is a series sparse linear filters with sizes of 5 × 5. It has a lower complexity and can be processed easily in the embedded processors of computers in real time. Experiments show that the Peak Signal to Noise Ratio(PSNR) of this algorithm is higher 4-6 db than that of the common bilinear algorithm. Furthermore, it effectively reduces the sawtooth phenomenon occurred in the interpolation algorithm and shows that its recover ability and real time property is superior to that of the traditional bilinear interpolation.