Accuracy Assessment of Haze Removal of SPOT 6/7 Imagery using Haze Index Algorithm Based on Machine Learning Classification

Multispectral optical remote sensing imagery in tropical regions such as Indonesia is often suffer from haze and cirrus contamination. This will affect to a reduction in image quality and accuracy of data interpretation. To mitigate these issues, a haze removal method has been developed employing the haze index algorithm which is based on reflectance values of the blue and red channels of SPOT 6/7 imagery. However the method must be tested to find out whether the result is sufficient for further implementation or not. The efficacy of the method was determined by testing the data with image interpretation using machine learning classification method to judge the efficiencies of the haze removal algorithm on SPOT 6/7 imagery. Experimental results indicate that the algorithm is proficient in information extraction using haze-free SPOT 6/7 data that has been processed using this method. As a result, data processed with haze removal based on the haze index technique can be useful for remote sensing applications.