Abstract: It is obvious that if a 2D image matrix is studied in a 1D space, any image pattern will take on stripe property in different degree. This stripe feature is popularly used in pattern discovery and pattern recognition. First, we use circle resampling to get 1D data from 2D image matrix. Second, we define a triangle base and its fractal IFS (iterative function system) and then a multi-freaquency analysis math model is constructed. In the end, we apply the model to different types of image analysis and image database discovery. The result indicates that the stripe feature obtained by the method is useful both in image recognition and pattern discovery.