Texture-Features-Based-Image-Retrieval-in-DWT-Domain
we have presented a Content Based Image Retrieval (CBIR) scheme using color, texture and shape feature information. Firstly an input RGB image is converted into YCbCr image and each part of the i.e. Y, Cb and Cr are extracted from it. Afterword, each components are uniformly quantized. Then BDIP and BVLC are computed over quantized Y-component on block size of 2ร2 and receive respective BDIP and BVLC image. Then on these two received image 3-level dwt is implemented and on each sub band some statistical parameters are evaluated to form a part of a feature vector. Now, on extracted quantized Cb and Cr components, 2-level dwt is performed and on each sub band some statistical parameters are calculated and this form second part of the feature vector. Now both parts are concatenated to form final feature vector. To proof that our system is adequate to retrieve good results, we have tested our scheme on benchmark database Coral-1000 . Same processed has been taken place for all the images present in the database and on the basis of the similarity measurement Top-20 results are retrieved and stored and the results are quite satisfying.
