Single Image Dehazing Using Global Illumination Compensation.

Area of Science:

• Computer Vision
• Image Processing

Background:

• Existing image dehazing algorithms struggle with background interference during atmospheric parameter estimation.
• This leads to suboptimal dehazing results and reduced image quality.

Purpose of the Study:

• To propose a novel global illumination compensation-based image-dehazing algorithm (GIC).
• To enhance the accuracy of atmospheric illumination and transmittance estimation by reducing background interference.

Main Methods:

• Illumination compensation in the CIELab color space using a shading partition enhancement mechanism.
• Computation of atmospheric illumination and transmittance parameters on enhanced images.
• Application of an improved atmospheric-scattering model for dehazing.

Main Results:

• The GIC algorithm demonstrated improved consistency in estimating atmospheric illumination and transmittance.
• Quantitative improvements were observed: Peak Signal-to-Noise Ratio (PSNR) increased by 3.25 and Structural Similarity (SSIM) by 0.084 compared to existing methods.
• Dehazed images exhibited reduced background interference.

Conclusions:

• The proposed GIC algorithm effectively addresses background interference in image dehazing.
• GIC offers superior performance over established methods, yielding more realistic and visually appealing dehazed images.