Rotation-invariant histograms of oriented gradients for local patch robust representation

Our research focuses on the question of feature descriptors for robust effective computing, presenting a novel feature representation method-rotation-invariant histograms of oriented gradients (Ri-HOG). Most of the existing HOG techniques are computed on a dense grid of uniformly-spaced cells and use overlapping local contrast of rectangular blocks for normalization. However, we adopt annular spatial bins type cells and apply radial gradient transform (RGT) to attain gradient binning invariance for feature descriptors. In such way, it significantly enhances HOG with respect to rotation-invariant ability and feature descripting accuracy. In experiments, the proposed method adopts object recognition as a test case and it is evaluated on PASCAL VOC 2007 dataset. The experimental results demonstrate that the proposed method is much more efficient than the existing methods.