A Space Variant Gradient Based Corner Detector for Sparse Omnidirectional Images

Omnidirectional cameras are useful in applications requiring rapid capture of image data representing the complete local environment. Feature detection from such image data is thus a prominent research issue. Transforming an omnidirectional image to a panoramic image may result in a sparse panoramic image with missing image data. Whilst image reconstruction techniques have been developed that enable the subsequent use of standard image processing algorithms, the development of image processing algorithms that can be applied directly to sparse image data has received less attention. We address the problem of corner point detection for sparse panoramic images by developing an algorithmic approach that can be applied directly to sparse unwarped omnidirectional images without the requirement of image reconstruction, and we illustrate the accurate performance of the algorithm through visual results and receiver operating characteristic curves.