Noise is conventionally considered undesirable in any system. Stochastic resonance (SR), on the other hand, is a counter-intuitive phenomenon, where addition of noise can be utilized to enhance the sensitivity of a non-linear system towards a weak subthreshold signal. In the past two decades, stochastic resonance, in both its dynamic and non-dynamic forms, has been utilized in a wide range of signal and image processing applications. The talk would present the speaker’s research contributions in the area of noise-aided image enhancement using dynamic stochastic resonance. In a physical system, dynamic stochastic resonance may be exhibited through the motion dynamics of a particle oscillating in a bistable double-well potential system in the presence of weak signal and tunable noise. An analogy of this bistable system in the context of image enhancement is presented, and an iterative equation, derived from the motion dynamics of the particle, is used for contrast enhancement of dark images in various domains. The term model, in the context of this talk, refers to the iterative processing equation and its parameter selection. Parameters for contrast enhancement are chosen from input statistics, and this model is, thus, referred to as the input statistics-dependent SR (ISSR) model. This model produces noteworthy contrast enhancement of dark images in various domains, but is unable to enhance images with both bright and dark regions. The parameter selection also suffers partial experimental dependence. To address the limitations of the ISSR model, in terms of derivation and application, a new intensity-specific value-dependent SR (IVSR) model is proposed to produce a quality of dynamic range compression in image. This model operates on a revised iterative equation, and produces a non-linear mapping of intensities such that the dark regions of an image (that has both bright and dark regions) are enhanced, while the already bright areas are preserved. The talk would be concluded with some potential applications and future research plan.
Dr. Rajlaxmi Chouhan received her Bachelor's degree in Electronics and Communication Engineering from Jabalpur Engineering College in 2009, and M.Tech. from IIITDM Jabalpur in December 2011. She received her PhD in August 2015 from the Department of Electronics and Electrical Communication Engineering at IIT Kharagpur. Dr. Chouhan served as the Chair of IEEE Women in Engineering Affinity Group of Kharagpur Section in 2014, and currently serves as a reviewer for various international journals, such as IEEE Transactions on Image Processing, Elsevier's Computers and Electrical Engineering, and a few others. Her research interests include noise-enhanced image processing, particularly image enhancement, image watermarking, image denoising, and applications of stochastic resonance. Her M.Tech. Thesis won the National Award for Best M.Tech. Thesis in Electronics and Electrical Engineering in 2012 from the Indian Society of Technical Education, New Delhi. She was also awarded the IEEE Region 10 (Asia Pacific) WIE Student Volunteer Award in 2014. More recently, her paper on noise-aided image enhancement has won the 2015 Premium Award for Best Paper in IET Image Processing.