¹School of Computer, Wuhan Textile University, Wuhan, China
²Computer School of Wuhan University, Wuhan, China

---

Abstract—Low frequency wavelet coefficients include much important vision information of the image, and are useful to image recognition and understanding. While at present, the applications of the low frequency wavelet coefficients are limited in the researches of image analysis. In this paper, the authors extracted the BFV (Binary Feature Vector, BFV) and TFV (Ternary Feature Vector, TFV) of low frequency wavelet coefficients based on non-parameter local transformation, which adopts the comparison results of the coefficient amplitudes in the neighborhood with the center coefficient to extract the feature rapidly. The TFV describes the texture more accuracy than BFV by adopting the two adaptive thresholds, and the by adjusting the parameter f TFV can adapt itself to the different texture data. The authors apply the BFV and TFV in aerial textures retrieval. In the experiments, our method is compared with the GLCM, Markov and Fractal algorithms, and the results prove that our methods behave well in the retrieval rate, especially the rapid processing speed.

Index Terms—Non-parameter local transformation, low frequency, wavelet coefficient, texture retrieval

[PDF]

Cite: Xubing Zhang, Bo Cai, "Non-Parameter Local Transformation of Low Frequency Wavelet Coefficients Applied in Aerial Texture Retrieval," Journal of Software vol. 6, no. 7, pp. 1313-1320, 2011.