Robust thermal bias estimation in real-world infrared images via integrating outlier suppression and second-order ℓ 0 norm sparsity

Removing the complex thermal bias from real-world infrared (IR) images degraded by thermal radiation effects is significantly crucial to improving image quality of high-precision IR cameras. Traditional IR image correction methods always neglect the impact of both unavoidable interference and salient image structures not to produce a high-contrast correction image. In this study, a robust and efficient thermal bias estimation method is proposed to compensate the complicated nonuniformity in noisy IR images. Specifically, the suppression strategy of outliers and salient structures is successfully made to significantly improve the estimation precision on the basis of the salient gradient difference between thermal bias and image structures. Meanwhile, the second-order ℓ 0 norm sparsity in gradient domain helps not only to avoid the staircase effects, but also effectively suppress system noise. The experimental results on real-world infrared datasets demonstrate the superiority of the proposed method over existing correction methods in terms of both correction residuals and visual quality