Smart Computational Imaging (SCI) Lab


Advanced Imaging Processing for Infrared Camera

Infrared focal plane array (IRFPA) sensors are widely used in the fields of aviation, industry, agriculture, medicine, and scientific research. However, the nonuniformity, produced by mismatches during the fabrication process of the IRFPA can considerably degrade the spatial resolution and temperature resolvability, since it results in a fixed pattern noise (FPN) that is superimposed on the observed images. Therefore, nonuniformity correction (NUC), being an indispensable key step, is applied to nearly all of the IRFPA-based engineering applications. Further, what makes the problem worse is that the nonuniformity varies over time and is closely related to external conditions, which results in the failure of traditional reference-based NUC methods. In order to solve this problem, scene-based nonuniformity correction (SBNUC) must be applied.

On the other hand, the raw data produced by modern high-quality infrared cameras have a wide dynamic range, which is often at the 12- to 14-bit level that typically exceeds the 8-bit sensitivity of a state-of-the-art display device. Furthermore, a human observer can distinguish only about 128 levels of gray (7-bit) in an image.1 Hence, a procedure aimed at reducing the data range must take place in the processing stage to enable the display to work with data from the detector. This procedure must accomplish two goals: reduce the dynamic range of the input image into a low range one that is acceptable for the display system and do this in such a manner that the output image is pleasing to the human observer.


Scene-based Nonuniformity Correction (SBNUC)
We are developing several simple and effective scene-based nonuniformity correction (SBNUC) techniques to adaptively compensate the FPN according to the scene information (no calibration targets). Our approaches are featured by fast convergence rate (less than 20 frames or even only use 2 frames), high steady-state accuracy, robustness to "ghosting artifacts" and computational simplicity. Those methods can be integrated with the hardwares of the infrared cameras to make them more robust to the temporal drift of  the FPN which may occur slowly over time. With those SBNUC technologies, we have successfully developed several ‘smart' infrared cameras, including two cooled thermal imager [3-5μm, HgCdTe IRFPA] with resolution 320×256 and 640×512, respectively; and two uncooled thermal imager with resolution 320×240 and 640×480 (8-14μm, VOx Microbolometer).


HDR IR Image Display and Detail Enhancement (DDE)
Dynamic range compression is an indispensable procedure for modern IRFPA imagers which has been known to be of great dynamic range compared to the current display devices. Furthermore, dynamic range compression, detail enhancement, and noise reduction are three mutually contradictory aspects but must be taken into consideration simultaneously. We have developed a new display and detail enhancement method for HDR IR images  that enhances the perceptibility of small details, and meanwhile avoid enhancing noises without causing artifacts.


Range Limited Bi-Histogram Equalization (RLBHE)

Histogram equalization is a popular technique for enhancing image contrast. However, it tends to change the brightness of an image and hence leads to annoying artifacts. We have proposed a novel extension of histogram equalization referred to as Range Limited Bi-Histogram Equalization (RLBHE). First, RLBHE divides the input histogram into two independent sub-histograms by a threshold that minimizes the intra-class variance. This is done in order to effectively separate the objects from the background. Then, range of the equalized image is calculated to yield minimum absolute mean brightness error between the original image and the equalized one. The proposed RLBHE can achieve visually more pleasing contrast enhancement while maintaining the input brightness.

SCILab 官方公众号
SCILab 官方B站