DRINK: Discrete Robust Invariant Keypoints

Computing descriptors for image features is a crucial task in many applications. A good feature descriptor is capable of providing invariance to geometric and lightning transformations while consuming as few memory as possible. Recently, there were proposed new approaches to compute a feature descriptor that rely on pixel intensity comparisons in order to generate a binary string, generating binary descriptors. However, binary descriptors only store a single bit per pixel comparison, and an useful portion of information, about how large the difference of intensity is, is lost due to this quantization. This work proposes a generalization of the binary descriptor idea: the discrete descriptor called DRINK. Using this idea, we are able to use more information related to the intensity difference while preserving the speed of the original binary descriptor. Our experiments show that the results produced by DRINK have similar or better precision level than other widely used binary descriptors and it is more than 3 times faster than ORB and about 20% faster than FREAK, while spending half of its bits to store a descriptor.