Six rice reflectance spectra in LOPEX 93 database were used as samples for decomposition into 10-layer signals based on the one-dimensional discrete wavelet types HAAR

DB4 and SYM4.The goals of this work were to reconstruct the signal through wavelet approximation coefficients and to calculate the wavelet fractal dimension of the reconstruction signal using the walking divider method.On each scale

we have determined the wavelet fractal dimension

wavelet detail coefficient variance

wavelet detail coefficient entropy

and approximate wavelet coefficient reconstruction curve variance.The results show fractal characteristics present in rice spectra

and proved the validity of fractal calculation by correlation coefficients greater than 0.9.The four parameters revealed that the turning point of the rice spectral characteristic scale is present on the sixth scale when rice spectral r esolution is less than 64 nm

in order to better reflect spectral peak-valley specific features.Through the measurement of 18 rice spectra in the field

this conclusion is further evidenced by two kinds of vegetation indexes and the correlation coefficients of two kinds of vegetation indexes and chlorophyll values on each scale.