Application of affine transformation to autoradiographic image compression using an artificial neural network
Date of Award
Doctor of Philosophy (Ph.D.)
Electrical and Computer Engineering
First Committee Member
Tzay Y. Young - Committee Chair
Second Committee Member
Weizhao Zhao - Committee Member
Autoradiographic imaging technique is a powerful approach to study local cerebral blood flow and glucose utilization in neuroscience research. A large number of autoradiographic image sequences of rat and mouse brains subjected to different experimental conditions have been generated in stroke or brain trauma research experiments. A large storage space is required to save these many autoradiographic image files. Data compression for autoradiographic image files is highly desirable. Following an established model that describes the z-direction evolution of sequential brain sections by affine transformation, this dissertation presents a novel lossless image compression method that reduces total entropy of sequential images dramatically by means of differentiation. A Hopfield neural network is designed to solve for the parameters of such an affine transformation. These parameters will be used to register autoradiographic image sequences. The difference image generated by subtracting registered images yields significantly reduced entropy, which is the core development in this dissertation. Detailed computer simulations have been performed for image registration, artificial neural network, and image data compression. An autoradiographic image sequence including seventy two rat brain images from a neurotrauma experiment has been processed by this method and other published methods. Comparison between results generated by different methods indicates that sequential image compression by the method presented in this dissertation has the lowest entropy. This method can also be further applied to other imaging modalities.
Engineering, Biomedical; Engineering, Electronics and Electrical
Riyamongkol, Panomkhawn, "Application of affine transformation to autoradiographic image compression using an artificial neural network" (2003). Dissertations from ProQuest. 2000.