Study of Dissociated Dipoles for Content Based Image Retrieval Using Stamp Recognition Systems

This paper proposes the use of soft dipole vectors for the purpose of Content Based Image Retrieval (CBIR). This paper illustrates the algorithm to evaluate soft dipole vector for a given image and as an example, depict its application in 'Stamp Recognition using CBIR'. Its advantages are its inherent rotation invariance and its uniqueness for every image, making it very robust. More importantly, under the section of "Rank-Proximity" we discuss about the unique behavior of soft dipoles which enables us to categorize images based on the soft dipole histogram values linking it to the physical attributes of that image. This direct link between the visual features and the mathematical model developed provide for a deeper and more intuitive understanding of soft dipole histogram in CBIR.