SIFT has a wide range of applications in the field of computer vision as a classic image matching algorithm, however, it is not robust enough on some optical and geometric transformations only by using the gradient histogram to build descriptor;on the other hand, the dimension of the SIFTs descriptor is too high to solve the real-time problems.In this paper, we introduce a local invariant descriptors -GCCI: Gradient Combine Contract Intensity Descriptor.Given the different contributions of neighborhood to the feature points, we divide the neighborhood area around feature points into two areas--internal area and peripheral area, then further divide them, next describe the divided areas respectively with the use of gradient orientation information and the contrast intensity.Our experiments demonstrate that using GCCI-SIFT in an image retrieval application results in increased accuracy and faster matching on some optical and geometric transformations, because of its high matching accuracy and efficient computation, the GCCI-SIFT has the potential to be used in a number of real-time applications.