Abstract
This paper addresses nwvo issues hindering the advances in accurate imnage alignment.First,the performance of descriptor-based approaches to image alignment relies on the chosen descriptor, but the optimal descriptor typically varies from image to image,or even pixel to pixel.Sec-ond,the neighborhood structure for smoothness enforce-ment is ustally predefined before alignmen.However,ob-ject boundaries are ofien better discovered during align-ment.The proposed approach tackles the two issues by adaptive descriptor selection and dynamic neighborhood construction.Specifically;we associate each pixel to be aligned with an affine transformation,and integrate the learning of the pirel-specific transfornations into image alignmen.The transformations serve as the common do-main for descriptor fitsion,since the local consensus of each descriptor can be esrimated by accessing the corresponding affine transformation.It allows us to pick the most plausi-ble descriptor for aligning each pixel.On the other hand,more object-aware neighiborhoods can be produced by ref-erencing the consistency between the leamed affine trans-formnations of neighboring pixels.The promising results on popular image alignment benchmarks manifests the effec-tiveness of our approach.