Visual Angular Haze Detection Using Focus Metrics

This paper proposes a digital image processing approach for measuring the visual angular haze of two kinds of glass developed by the American Glass Products Company (AGP) group: a low angular haze glass and a standard one. This approach applies eight focus metrics, commonly used in autofocus algorithms, on natural images affected by each type of glasses. Then, evaluate the sharpness of each image and find a relation between the visual angular haze and the sharpness given by each metric. In the process, the camera was rotated from 40 to 90 ^∘, in steps of 10 ^∘, to obtain a range of different angles, where the latter represents the front of the scenery. Subsequently, a region of interest (ROI) is manually selected in the area of each image with the highest blur, and each focus metric is applied to the ROI. A linear regression model to approximate the haze level is obtained for each glass with a respective weight for each metric used. The results of these experimental linear relations give a coefficient of determination values up to 0.9996, which indicates they fit with the empirical data after selecting the right combination of five focus metrics.