Objective haze removal assessment based (Two-objective optimization

Abstract

Recently, single image haze removal has been investigated extensively. However, its performance is evaluated mainly based on subjective visual quality of recovered images. Several objective assessments have been attempted to evaluate haze removal schemes objectively, such as no-reference image quality model and edge-based assessments. Unfortunately, these objective assessments are not able to measure visual quality of recovered images appropriately, since the visual quality of recovered images is not concerned. Note that an objective assessment for haze removal should consider dehazing effect and distortions introduced during the haze removal process. In this paper, an objective assessment for haze removal based on two-objective optimization is presented where both dehazing effect and distortions of color and artifacts in recovered images are considered. The two indicators are combined with a weighted Euclidean distance. Two examples with two dehazing schemes are provided to justify the proposed objective assessment for haze removal. The simulation indicates that the objective results are consistent with their subjective visual quality of recovered images. It suggests that the proposed objective assessment may be applied to evaluate different haze removal schemes objectively.