Abstract: In this paper, an image wavelet transform is conducted by using the features of
Daubechies orthogonal wavelet and the Mallat multi-scale analysis method. The image is
decomposed into four parts of a low-frequency contour, the horizontal high-frequency, the vertical
high-frequency and the slash high frequency. Because the image edge mainly concentrates in the
high-frequency parts, the high-frequency wavelet coefficients are reserved while low-frequency
wavelet coefficients are conducted of wavelet transform again, extracting the second
high-frequency signals from low-frequency contour parts. Finally, the maximization of edge
information of image is achieved in extraction by use of inverse transformation of the
high-frequency wavelet coefficients and the second high-frequency wavelet coefficients.

Key words: Daubechies wavelet, wavelet transform, multi-scale analysis, edge detection