Studies And Applications Of Wavelet Analysis In Image Processing

In this thesis, we study the traditional wavelet and multi-band wavelet analysis theory and their applications in three fields. The three fields are: Iris recognition, content-based image retrieval and image coding.In iris recognition field, an iris recognition method based on 2D Gabor wavelet and RBF Neural Networks is put forward. The iris features are extracted through 2D Gabor wavelet in different directions and different frequencies. And a RBF Neural Network is designed as a classifier. The experimental results show that the method is resonable with a good performance. In content-based image retrieval field, a multi-band wavelet analysis based image retrieval algorithm is proposed. The features of the images are extracted after performing a 4-band wavelet transform. And the images are represented as a vector with length of 48 decimal fractions. The image retrieval is conducted the Euclid distance between the vector of input image and the ones in the images database. The experimental results and conclusion prove that the algorithm in this thesis have high accuracy and calculation speed. In image coding field, we bring forth a multi-band wavelet analysis based image coding method. In our algorithm, the first step is to decompose the image into three channels of data such as RGB or YCbCr, then perform the multi-band wavelet transform in each channel, then quantitate and organize the coefficients, at last write the data to a bit stream file after entropy coding. In order to suit for our bit stream file format, we choose run-length coding method to encode the coefficients. The ROI (Region Of Interest) coding is actualized by our algorithm.With the three instances in image processing field, we discuss the applications and algorithms of wavelet analysis in image processing detailedly. The experimental results and conclusions show that wavelet analysis is a powerful tool for depicting the relativity structure of image signals, and is useful in data compression and signal approach.