Stripe non-uniformity correction of infrared images using parameter estimation

Abstract

Non-uniformity of the readout circuits in the infrared focal plane array (IRFPA) results in stripe noise in the infrared image, which badly degrades the visual quality and the usability of image. The existing destriping methods typically try to directly restore the clean image from the noisy one. However, it may not be a good strategy because it is hard to distinguish detail information and stripe noise in the image. In this paper, the destriping problem is formulated as a parameter estimation problem according to the local linear relationship between the clean image and the stripe noise. To estimate the stripe noise parameters, an optimization model is proposed based on the assumption of the smoothness of the clean image and the random characteristics of stripe noise parameters. The optimization problem can be solved using the gradient descent method. Then, an objective quality metric referred as DS-SSIM is proposed to evaluate the performance of the destriping methods. The qualitative and quantitative comparison with some state-of-the-art methods on the infrared images with simulated and real stripe noise demonstrate the advantages of the proposed method.