Luminance Component Analysis for Exposure Correction

TL;DR

Inspired by principal component analysis (PCA), an exposure correction method called luminance component analysis (LCA), which applies the orthogonal constraint to a U-Net structure to decouple luminance-related and luminance-unrelated features.

Abstract

Exposure correction methods aim to adjust the luminance while maintaining other luminance-unrelated information. However, current exposure correction methods have difficulty in fully separating luminance-related and luminance-unrelated components, leading to distortions in color, loss of detail, and requiring extra restoration procedures. Inspired by principal component analysis (PCA), this paper proposes an exposure correction method called luminance component analysis (LCA). LCA applies the orthogonal constraint to a U-Net structure to decouple luminance-related and luminance-unrelated features. With decoupled luminance-related features, LCA adjusts only the luminance-related components while keeping the luminance-unrelated components unchanged. To optimize the orthogonal constraint problem, LCA employs a geometric optimization algorithm, which converts the constrained problem in Euclidean space to an unconstrained problem in orthogonal Stiefel manifolds. Extensive experiments show that LCA can decouple the luminance feature from the RGB color space. Moreover, LCA achieves the best PSNR (21.33) and SSIM (0.88) in the exposure correction dataset with 28.72 FPS.

Figures (11)

• Figure 1: The main idea of LCA. In the training phase the luminance information is first extracted from two identical images with different exposure values, then the luminance-related and luminance-unrelated components are decoupled by orthogonal decomposition. During the inference phase, the well-trained decomposition modules are directly applied to the input image without the ground truth.
• Figure 2: The structure of OLD.
• Figure 3: Toy problem. Subtracting two identical images with different exposure reveals luminance differences (representing luminance information). Luminance information is coupled in the original RGB space, as indicated by high R, G, and B component variance.
• Figure 4: Decoupling the luminance-related feature by different methods. Higher variance indicates the more luminance information the corresponding component contains.
• Figure 5: Illustration of the geometric optimization process.