Resolving Implementation Ambiguity and Improving SURF
This work addresses implementation ambiguities in SURF for computer vision practitioners, but it is incremental as it builds on existing SURF methods.
The paper tackled the problem of inconsistent performance and algorithmic details in SURF implementations by defining techniques for feature stability and proposing improvements for speed and stability, resulting in an implementation that is several times faster and more stable.
Speeded Up Robust Features (SURF) has emerged as one of the more popular feature descriptors and detectors in recent years. Performance and algorithmic details vary widely between implementations due to SURF's complexity and ambiguities found in its description. To resolve these ambiguities, a set of general techniques for feature stability is defined based on the smoothness rule. Additional improvements to SURF are proposed for speed and stability. To illustrate the importance of these implementation details, a performance study of popular SURF implementations is done. By utilizing all the suggested improvements, it is possible to create a SURF implementation that is several times faster and more stable.