Model-based adaptive enhancement of far infrared image sequences Ming Tang * , SongDe Ma, Jing Xiao National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, P.O. Box 2728, Beijing 100080, People’s Republic of China Received 4 June 1999; received in revised form 20 March 2000 Abstract Two enhancement algorithms, the spatial and spatiotemporal homomorphic filtering (SHF and STHF), were proposed in (Highnam, R., Brady, M., 1997. IEEE Trans. Pattern Anal. Machine Intell. 19, 410–415) to enhance far infrared images based upon a far infrared imaging model. In this paper, we theoretically and experimentally prove that the enhanced results with SHF are in general smoother than those with STHF, although STHF may reduce the processing time greatly in comparison to SHF. Based on this conclusion, an adaptive spatiotemporal homomorphic filtering (ASTHF) algorithm is proposed. The adaptive factor of ASTHF is also discussed in detail to obtain the trade- o between smoothness and convergence time. Ó 2000 Elsevier Science B.V. All rights reserved. Keywords: Image enhancement; Far infrared imagery; Adaptive filtering; Homomorphic filtering; Spatiotemporal filtering; Prometheus 1. Introduction It is dicult in some situations such as foggy days to obtain dependable information from the visible part of the electromagnetic spectrum be- cause sensing in that part deteriorates badly. In such cases, sensing in the far infrared part may be superior. To reduce the cost, however, an uncooled imaging device may have to be used. Conse- quently, one is confronted by images with rela- tively poor SNR. Conventionally, image enhancement is per- formed to remove noise and to improve SNR by applying general routines which seldom consider the actual modality involved. Therefore, it is hard to ensure that important features are pre- served, and that artifacts do not appear. More- over, if the images are consecutive frames of an image sequence the enhancement of a single isolated image ignores the interframe temporal information. Based on the above consideration, Highnam and Brady (1997) analyzed carefully the far infra- red imaging process and the factors of image de- grading, and showed that an ideal far infrared image would consist solely of emission and would be piecewise constant. Then they designed two algorithms, the spatial and spatiotemporal homo- morphic filters (SHF and STHF) to enhance far infrared images. STHF makes explicit use of the temporal information provided by the image www.elsevier.nl/locate/patrec Pattern Recognition Letters 21 (2000) 827–835 * Corresponding author. E-mail address: tangm@nlpr.ia.ac.cn (M. Tang). 0167-8655/00/$ - see front matter Ó 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 8 6 5 5 ( 0 0 ) 0 0 0 4 1 - 6