Spiking Cortical Model for Rotation and Scale Invariant Texture Retrieval

Retrieval of texture images, especially those with different orientation and scale changes, is a challenging and important problem in image analysis. This paper adopts spiking cortical model (SCM) to explore geometrical invariant texture retrieval schemes based on Discrete Cosine Transform (DCT) coefficients of pulse images. The series of pulse images, outputs of SCM, have a robust talent for extracting edge, segment and texture which are inherent in the original images, but they are large 2-dimensional image data so that it is difficult to process further. Geometrical invariant features of the original images can be extracted by characterizing the pulse images in DCT domain, which is dramatically reduced the large data to a small 1-dimensional vector. Many experiments and comparative studies are performed to show that the retrieval schemes are novel and effective in extracting invariant features.