Weixian QIAN^*, Qian CHEN, and Guohua GU

440 Lab, EEOT, Nanjing University of Science and Technology, Nanjing, Jiangsu Province 210094, China

(Received July 19, 2009; Accepted November 6, 2009)

Nowadays, almost all scene-based nonuniformity correction (NUC) algorithms have the same shortcoming: a low convergence speed. This shortcoming produces ghosting artifacts. In this paper, we will start by discussing the traditional temporal high-pass NUC algorithm, and then combine the space frequency and temporal frequency. This combination will produce a new algorithm called the space low-pass and temporal high-pass NUC algorithm. The kernel idea of this algorithm is to eliminate the nonuniformity's high-space-frequency part and retain the nonuniformity's low-space-frequency part. Moreover, the high-space-frequency part's processing is controllable, which markedly increases convergence speed. These features make our algorithm's convergence speed so high that processed images have almost no ghosting artifacts.

Key words: scene-based nonuniformity correction, ghosting artifacts, temporal high-pass, space low-pass, convergence speed

^*E-mail address: developer_plus@163.com