Wavelet based Chordiogram Image Descriptor for Image Retrieval

In this work, an efficient multiresolution approach based geometric details of localized edgels is proposed to represent an image, called wavelet based chordiogram image descriptor. The proposed approach mainly focuses on reducing the computational cost of retrieval system with the preservation of accuracy and resilient to illumination, scaling, noise and small rotations, and is performed by deriving the geometric features from the pyramid structure of an image using Discrete Haar Wavelet Transform. The proposed multiresolution approach based chordiogram image descriptor is comprehensively tested on benchmark datasets. The experimental results proved that proposed approach is considerably better in terms of time cost and retain the accuracy of existing approach.