My PhD research project focuses on the image retrieval in a compressed domain. A new content-based image retrieval technique using prediction error concept called CBIR-PE system has been developed. The proposed system makes use of a new clustering scheme employing wavelet based contourlet transform and fuzzy c-means clustering for improving the performance of the system. A multilayer neural network predictor has been employed for predicting the pixel values using the neighboring pixels.
The proposed method has the advantages of using lower bandwidth for image transmission and reduced storage space for image storage since only the error values instead of the actual pixel values are used for transmission and storage. In addition, it also has the advantage of data security since the original images cannot be reconstructed without the predictor parameters. Experimental results with medical and other image databases have confirmed the superiority of the proposed method over the other existing methods.
(Won Best PhD Thesis Award 2018, Faculty of Computing and Informatics, Multimedia University, Malaysia)