Performance Analysis of Histogram Equalization and Color Balance for Underwater Image Restoration

Abstract

Underwater images commonly suffer from color casts, low contrast, and reduced visibility caused by wavelength-dependent absorption and scattering of light in water. Histogram Equalization (HE) and Color Balance are two classical spatial-domain techniques frequently applied to mitigate these degradations, yet their comparative performance across different underwater degradation conditions is rarely quantified beyond visual inspection. This study analyzes the performance of per-channel HE and the Simplest Color Balance algorithm on nine real underwater images grouped into three degradation categories: blue-dominant, green-dominant, and dark/turbid. Objective evaluation used Mean Squared Error (MSE), Peak Signal-to-Noise Ratio (PSNR), Shannon entropy, RMS contrast, and a color-cast index, complemented by qualitative visual comparison. Color Balance consistently achieved higher PSNR and lower MSE than HE across all nine images (average PSNR of 16.12, 18.36, and 14.69 dB for the blue, green, and turbid categories versus 11.00, 13.66, and 10.54 dB for HE), indicating more faithful restoration with fewer artifacts. HE produced a greater increase in entropy and contrast and reduced the numerical color-cast index more aggressively, but amplified noise and introduced reddish/magenta artifacts, particularly on the darkest, most compressed image, where PSNR fell to 7.17 dB. Both methods performed best on moderately degraded blue- and green-dominant images and struggled on severely dark, low-bit-depth images, revealing a clear capability boundary for classical spatial-domain restoration. These findings offer practical guidance for selecting or combining HE and Color Balance in underwater image processing pipelines.