why use GLCM for feature extraction

Why is it necessary to use GLCM or other such methods when finding image features? GLCM matrix is found from the original image matrix so how come the properties such as energy, correlation etc found from the original image do not depict the true character of the image while GLCM matrix does?