One Improved OpponentSIFT and HOPDFG Descriptor Based Image Classification Algorithm

There are abundance of image classification and recognition algorithms, which Bag of Features (BoF) model based ones are the most classic and effective methods. In this article, I use an original, high-performance characteristic elicitation strategy for picture classification in the methodological framework of the BoF approach. Taking into account the overall theoretical and experimental results, I use the OpponentSIFT which is invariant to photometric transformations such as ray intensity changes and ray colour changes. Also, I enhance the Histogram of Oriented Gradients (HOG) descriptor by subjecting the image to the simple yet powerful probability density function (p.d.f) and then calculating HOPDFG from the outcome in a manner that abounds the form information encoded in the descriptor. In the experiments, I compare the picture classification performance of the method used and related fusion techniques with other existing methods on various datasets and the method used exhibits superior performances on object recognition and scene categorization.