ABSTRACT

In this thesis performance of image matching techniques is analyzed. Here two existing techniques i.e. scale-invariant feature transform (SIFT) and Speeded up robust features (SURF) are used to develop a HAECBIR technique. Performance of these three technique is analyzed by using Oxford building dataset. The data set consist a set of different image classes. These techniques are used to extract visual features of images. Frequency vector is recorded a Histogram for spatial tile of the image. The resulting vector are used for feature matching with other images the performance with respect to accuracy and time is evaluated for these three techniques. The result Indicate that matching accuracy of the Hybrid technique is better than SIFT and SURF. The hybrid approach is also faster than SIFT in terms of execution time to retrieve similar images from dataset.
Keywords: Feature Extraction, A hybrid Approach Efficient Content based Image retrieval system (HAECBR), Speeded up robust features (SURF), Scale Invariant feature transform (SIFT)