Primal sketch of images based on empirical mode decomposition and Hough transform

Primal sketch is a sparse representation for image. It is a difficult task in low-level computer vision. In this paper, an image primal sketch method is presented by using empirical mode decomposition and Hough transform. The key of image primal sketch includes two steps. One is extracting features from image. The other is representing the image features. Using empirical mode decomposition technique in features extracting process decomposes the gradient magnitude image. The intrinsic mode functions are obtained and they catch information under different scales of image, where the first intrinsic mode function represents the image information under finest scale of image. In features expressing process, the sum image of the first and the second intrinsic mode function is segment by a threshold and the binary image that representing the features of original image can be obtained. Using line segments detected by Hough transform in the binary image draws the primal sketch. Experiment results demonstrate that the primal sketch of image regarding line segments as its primitive can reflect image features effectively.