Application of Non-Linear Encoding to Picture Transmission

Abstract

The process of converting nonlinear analog signals to linear digital signals is a type of companding. This process of companding consists of compressing or expanding the dynamic range at the transmitter and restoring the original levels at the receiver. In telephony it is used to account for differences in speakers' voices. A loud voice will not overdrive the channel, yet a soft voice can be heard. In image transmission and processing, companding is even more important because of the nature of image forming. Both natural and photographic image formation are multiplicative processes. In a natural scene, the illumination and reflectance of objects are combined by multiplication to form observable brightness. Since this combining is a nonlinear process, it is important to transform the output to a linear signal at the earliest possible point in the transmission. If linearizing is not done, noise will affect the dark portion of the picture much more than the bright areas. Companding can be accomplished in many ways either by analog or digital method. The most common analog method is the use of log amplifiers with nonlinear amplitude gain. The most common digital technique is nonlinear encoding which performs the companding while the analog signal is being converted to digits. This companding process, when used on the output of a photo scanner, can be used to improve the transmission and reconstruction of digital pictures.