Abstract
In recent years, image indexing techniques have become important with the rapid growth of digital images and video libraries. As the volume of image information grows, the management tools must become increasingly reliable. A two steps approach to search the image database is adopted. First for each image in the database, a feature vector characterizing some image properties is computed. Second, given a query image, its feature vector is computed too, compared to the feature vectors stored in the database, and try to find out the most similar images to the query image.Our work combines color and texture for image retrieval, and we allow user interaction. Wavelet transform has become more and more popular, some standards like JPEG 2000 is based upon wavelet transform instead of DCT. A new complex wavelet transform gives a fast way to generate texture features.In our method, there are two steps to retrieve image . First, we segment the image into several clusters by the color using C-Means algorithm and apply complex wavelet transform to each cluster to compute the feature vector and store the feature vector in out feature database. Second, we apply the same process to our query image and compute the feature vector. Then we compare the feature vector with those stored in the feature database to find out the most similar images.