Efficient object matching using partial dominant orientation descriptor
Abdessamad Elboushaki, Rachida Hannane, Karim Afdel, Lahcen Koutti · 2015
This paper presents a method for extracting a new feature descriptor, named Partial Dominant Orientation Descriptor (PDOD), which can be used to perform reliable matching between objects with similar as well as different textures. The object is represented by a set of key locations of stable points using Difference of Gaussian, so that the matching can proceed successfully despite changes in viewpoint, scale, illumination, and distortions. The proposed descriptor at feature point takes into account the position and partially computes the dominant orientations of other key locations relative to this point, thus, offering a global discriminative characterization. The correspondence between two objects is performed by matching each key location in the first object independently to the set of key locations extracted from the second object using Euclidian distance. The one with the smallest distance is picked as the best candidate match. The descriptor is highly distinctive, and provides robust matching across a substantial range of rotation variance, change in textures, and object deformation. Our experiments on KONKLAB public dataset indicate that our method outperforms other benchmarks such as SIFT, PCA-SIFT and SURF matching algorithms.