Feature Extraction of Low Frequency Wavelet Coefficients Based on Non-Parameter Local Transformation

Low frequency wavelet coefficients are very useful to image recognition and understanding. While the applications of the low frequency wavelet coefficients are limited in nowadays. In this paper, the authors present a new method of the image retrieval by extracting BFV (Binary Feature Vector, BFV) and TFV (Ternary Feature Vector, TFV) of low frequency wavelet coefficients based on non-parameter local transformation, which develops the application of the low frequency wavelet coefficients. On the other hand, TFV features extraction method overcomes the disadvantages of the typical census transformation by using the adaptive threshold and the adjustive coefficient f. In our experiments, our method is compared with the GLCM, Markov and Fractal algorithms, and the results prove that our method is feasible and effective.