Visual Angular Haze Detection Using SSIM

Delgado C.; Castro O.M.; Giuffra A.; Murray V. · Smart Innovation, Systems and Technologies (2022) · DOI: 10.1007/978-3-031-08545-1_56

Resumen

In this work, we present an approach to calculate the haze of a Polymer Dispersed Liquid Crystal (PDLC) with respect to the viewing angle of the object using the Structural Similarity Index Measure (SSIM) method. The method shown in this project is developed based on a limited but authentic image dataset provided by an international glass company. This method calculates the measure of haze, dividing the input image into the same size regions and extracts the ones where the image does not have haze, to use those as a reference against the frames with haze. Then, based on the SSIM components, the system produces a haze grid and calculates two features, the scope of linear regression and the mean of the grid. These features are used to obtain a value that represents the amount of haze in the glass. The final model is approximated using a quadratic regression with MSE = 5.388.

Instituciones: Universidad de Ingeniería y Tecnología