Non-Uniformity Correction for Infrared Images

In this article, we address the processing of non-uniformity noise in infrared images. This type of noise causes non-uniformity effects that significantly degrade image quality. To resolve this issue, we employ a Recursive Least Squares (RLS) algorithm based on a quadratic polynomial model. The algorithm implementation involves dynamically estimating correction parameters through adaptive filtering, where each pixel's non-uniformity is modeled using a second-order polynomial function of its coordinates. Key computational steps include: 1. Initializing covariance matrices and weight vectors for each detector element 2. Implementing real-time parameter updates using the RLS gain calculation 3. Applying polynomial-based correction coefficients to raw infrared data 4. Optimizing forgetting factors for balance between adaptation speed and stability Through this methodology, we achieve more accurate and clearer infrared images, enabling enhanced research analysis and practical applications. The code structure typically involves separate modules for calibration data collection, RLS parameter estimation, and pixel-wise correction application, with optimized matrix operations for efficient real-time performance.