Dimensionality reduction through PCA over SIFT and SURF descriptors

One of the constant challenges in image analysis is to improve the process for obtaining distinctive object characteristics. Feature descriptors usually demand high dimensionality to adequately represent the objects of interest. The higher the dimensionality, the greater the consumption of resources such as memory space and computational time. Scale-Invariant Feature Transform (SIFT) and Speeded Up Robust Features (SURF) present algorithms that, besides of detecting interest points accurately, extract well suited feature descriptors. The problem with these feature descriptors is their high dimensionality. There have been several works attempting to confront the curse of dimensionality over some of the developed descriptors. In this paper, we apply Principal Component Analysis (PCA) to reduce SIFT and SURF feature vectors in order to perform the task of having an accurate low-dimensional feature vector. We evaluate such low-dimensional feature vectors in a matching application, as well as their distinctiveness in image retrieval. Finally, the required resources in computational time and memory space to process the original descriptors are compared to those resources consumed by the new low-dimensional descriptors.