Kernel Descriptors for Visual Recognition
   Liefeng Bo, Xiaofeng Ren, and Dieter Fox, at NIPS 2010



Abstract

The design of low-level image features is critical for computer vision algorithms. Orientation histograms, such as those in SIFT and HOG, are the most successful and popular features for visual object and scene recognition. We high- light the kernel view of orientation histograms, and show that they are equivalent to a certain type of match kernels over image patches. This novel view allows us to design a family of kernel descriptors which provide a unified and princi- pled framework to turn pixel attributes (gradient, color, local binary pattern, etc.) into compact patch-level features. In particular, we introduce three types of match kernels to measure similarities between image patches, and construct compact low-dimensional kernel descriptors from these match kernels using kernel princi- pal component analysis (KPCA). Kernel descriptors are easy to design and can turn any type of pixel attribute into patch-level features. They outperform carefully tuned and sophisticated features including SIFT and deep belief net- works. We report superior performance on standard image classification bench- marks: Scene-15, Caltech-101, CIFAR10 and CIFAR10-ImageNet.