THEORY AND APPLICATIONS OF ABSOLUTE MOMENT BLOCK TRUNCATION CODING OF VIDEO DATA (QUANTIZATION, COMPRESSION, SOURCE CODING)
Abstract
A new image coding method using preservation of absolute moments as a design criterion is discussed along with applications to still and moving imagery. This method is called Absolute Moment Block Truncation Coding or AMBTC. Theoretical analysis is done to find the mean square error produced by this quantizer. Results show that for symmetric input histograms AMBTC is equal to the minimum mean square error non-parametric quantizer, whereas in the general case it is a suboptimum implementation. A threshold depending on the skewness of the input histogram is proposed to achieve a lower mean square error while keeping a non-parametric fixed threshold scheme. Theoretical analysis is also carried out to find the marginal probability density function at the output under the assumption of a Markov-Gaussian input sequence to the quantizer. Practical applications to moving imagery are discussed. A fixed rate video compression system is presented. The system uses motion compensated prediction of the parameters associated with AMBTC. Adaptive quantization depending on local variations is used in the transmission of one of the error signals. Medium quality video transmission is possible using the system with bit rates less than one bit per pixel.
Degree
Ph.D.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.