Fog detection using GLCM based features and SVM
Rakesh Asery, Ramesh Kumar Sunkaria, Lakhan Dev Sharma, Aman Kumar · 2016
Classification between foggy and non-foggy images is a primitive step for automation in traffic activity and industries. The existing techniques provide low accuracy and needs validation over both synthetic and natural database. Foggy images are identified and classified based on their optical characteristics for vision enhancement and to make them more efficient for further processing. In proposed work, Gray Level Co-occurrence Matrix (GLCM) features are extracted and significant features are selected using boxplot for classification between foggy-images and nonfoggy images. Three parameters, Contrast, Correlation, and Homogeneity are used as classification parameters for Support vector machine (SVM) classifier. These parameters are suitable for both synthetic as well as natural database. Results revealed that the proposed technique classifies between foggy and non-foggy images with high accuracy of 97.16% and 85% on synthetic and natural database respectively.