Many people nowadays can benefit from localization since the ubiquitous smartphones are integrated with many facilities such as visual sensors which can provide absolute or relative localization. Visual Fingerprinting (VFP) is an image matching technique which is based on visual information and provides localization to the user by finding the closest images in the database of images or videos. This approach has the flexibility to compensate between the accuracy and the practical implementation requirements. VFP especially becomes an attractive localization solution for the urban indoor environment due to certain number of facts such as limited access to the GNSS signals, the abundance of visually recognizable and distinguishable features in these environments, and that it does not require device calibration. Most VFP methods that are developed to be implemented on the smartphones include integration with other sensors such as Wi-Fi. These methods are focused mostly on the accuracy and less so on the computational aspects. They usually do not provide a detailed analysis that would suggest the feasibility of real-time implementation of their methods on smartphones. This study investigates the effectiveness of different VFP methods by evaluating the accuracy of the matching, preprocessing time and average matching time. To perform VFP, several things are required; a database of the location-tagged images, an algorithm to process the uploaded images; an algorithm to find the match in the image database and a localization algorithm that infers a location for the user based on the aforementioned steps. This research is focused on the first three steps. Explicitly the local detectors and descriptors have been used and their important properties have been studied. To further enhance the performance, multi-dimensional data structure has been deliberated. Consequently, a set of different combination for detectors/descriptors, data structures and dimensionality reduction algorithm has been chosen to be evaluated. The results obtained from the evaluations show that with the proper selection between these combinations, they can deliver the practical requirements to implement VFP-based localization on smartphones.