In order to ensure the reconstruction quality of an image and realize the function of real time processing, a single channel interpolation algorithm based on gradient is proposed. By introducing horizontal and vertical gradient of the Green channel, the edge information can be preserved better, and the color of the smooth area becomes more smooth simultaneously. And by simultaneously introducing the 45o and 135o gradients of the Blue and Red channels as the additional edge judgement, the edge information of all the three channels can get a sufficient judgement, so the edge of the whole image become more clearer. Then, the quality of the interpolation algorithm will be evaluated and verified by the MATLAB simulation. The experiment results show that, compared with Nearest Neighbor Replication and Bilinear interpolation, a better quality can be got not only on edge but in the smooth area, and the zipper effect is reduced apparently. Besides, the interpolation algorithm is designed for the real time processing under the precondition of ensuring the image reconstruction quality. Theoretical analysis indicates that the proposed algorithm is easy to realize for hardware wiithout complicated multiplication and division operation, and the procedure of the interpolation is suitable for parallel processing.