Texture Recognition Based on DCT and Curvelet Transform

Abstract

This paper presents a proposed technique for texture recognition which depends on the combination of Discrete Cosine Transform (DCT) with Fast Discrete Curvelet Transform (FDCvT) via Wrapping.The proposed technique includes two stages, the first stage is implemented by taking individual natural textures (wood, stone and grass) with several positions and calculation of the features vector (Mean and standard deviation) by using many methods: DCT, FDCvT via Wrapping, and both FDCvT via Wrapping and DCT. The second stage is implemented by taking several samples of new textures for testing the work.The results show that the texture recognition rate by the DCT is 52%, and the FDCvT via Wrapping is 88%. But the new technique of (FDCvT via Wrapping and DCT) achieves better recognition rate (92%). This combination leads to efficiency in texture recognition because the DCT added some qualities that strengthen the work of the Curvelet Transform.