Seamless cloning of a source image patch into a target image is an important and useful image editing operation. In order to achieve a balance between effect and timespace consuming, this paper applies both guided derivative and closed form intensity interpolation to minimize the disturbing edges and maximize the similarity between target and source images. The image is decomposed into four parts: the approximation parts, the horizonal, the vertical and the diagonal details by discrete wavelet transform. Approximation parts is processed by interpolating pixel values through smoothing the gradient domain and pixel values in other parts are got by direct assignment and linear interpolation. This combination is advantageous in terms of speed, small memory footprint, realizing real time cloning of large regions. We can reduce the time cost to 1/16 of Poisson cloning and achieve better effect for some extreme gradient. In order to define and get the best possible effect, this paper introduces a formal cost function for the evaluation of the quality of cloning. By analysing the cost function that represents the mixture effect we bring about in this paper, our method is proved quite effective in lots of situations.