A fixed-point local tone mapping operation for HDR images

摘要

This paper proposes a fixed-point local tone mapping operation (TMO) for high dynamic range (HDR) images. A TMO is classified in two types: local and global. Although a local TMO offers better results than global one, it requires more resources such as a computational cost and memory space. The proposed method uses fixed-point arithmetic with short data to solve this problem. The method uses an intermediate format which is composed of 8-bit mantissa part and 8-bit exponent part instead of IEEE754 standard floating-point format. Moreover, these mantissa part and exponent part are processed separately as two integer numbers. As a result, the method reduces the memory space. In addition, the method also reduces numerical range of calculations; it facilitates to implement the method with fixed-point arithmetic. The experimental results show that the method reduces the memory and computational costs, and offers high quality of tone mapped images comparable to the conventional method.