Research Of Image Retrieval Based On Dual-tree Complex Wavelet Transform And Information Entropy

Wavelet transform (WT) is known as the mathematical microscope about signal analysis. In the image feature extraction, especially Mallat pyramid algorithm (PDWT), is a good choice. But the real wavelet has some defects, such as orientation information less, time shift sensitivity and the lack of phase information. To overcome these shortcomings, multiwavelet and complex wavelets have been proposed.Complex Wavelet Transform (CWT) is a wavelet to expand complex domain, which not only keeps the good tradition of wavelet time-frequency localization characteristics, but also overcomes the shortcomings of real wavelet. Dual-tree Complex Wavelet Transform (DTCWT) not only retains the wavelet multiresolution characteristics and good time localization analysis ability, but also has the advantage of complex wavelet transform, such as good shift invariance, directional selectivity, and direction of limited data redundancy. At the same time, It can guarantee completely reconstruction through the filter.Aiming at improving the retrieval function of texture image, In this paper, a new image retrieval method was proposed based on the dual-tree wavelet and information entropy. According to the resolving features of complex wavelet, this thesis starts with the modulus energy of wavelet with the histogram distribution characteristics between the real component of high-frequency complex modulus and the imaginary part. Using the mathematics features of the image entropy and being measured by information relative entropy, the experiment combines the features sitilling and similitude computing or retrieval by means of the maximum likelihood estimation regulations. The experiment analysis proves that this method helps improve the image retrieval efficiency.