Modeling of overhead information for noise bias compensation
2023 年 11 巻 1 号 p. 1-7
詳細
2023 年 11 巻 1 号 p. 1-7
When acquiring night scene images, brighter images can be produced by increasing the ISO sensitivity. However, there is an associated problem with this approach, in that noise is also amplified by such brightness adjustment in digital cameras. Two noise components are amplified by this brightness adjustment: the noise bias (i.e., noise with a nonzero mean value) and the noise variance. Our research group is studying noise bias compensation (NBC), which is a new approach to noise reduction corresponding to the noise bias. Compensation values are calculated using pixel value information from the observed image and the clean image. In previous NBC methods, it was assumed that all pixel values in an image that contains no noise (clean image) were included in the overhead information. In this paper, we propose a method for compensation value calculation that does not rely on the overhead information for night scene images acquired with a general-purpose digital camera at high ISO sensitivity. Experimental results confirmed the effectiveness of the proposed method.
