On Using SIFT Descriptors for Image Parameter Evaluation

In this work we present a composite method for image parameter evaluation using Scale-Invariant Feature Transform (SIFT) descriptors and bag of words representation applied to pre-selected image parameters, with potential applications to solar data and other domains. As one of the main challenges in computer vision, image parameter evaluation has been approached from supervised and unsupervised perspectives. Taking advantage of the SIFT scale and rotation invariant properties; we propose a combined method that will aid the image parameter selection process when applying SIFT and bag of words on top of pre-selected parameters. We provide a comparison against traditional methods and across several different datasets to validate our method.