Tensor Voting for Image Correction by Global and Local Intensity Alignment
IEEE Transactions on Pattern Analysis and Machine Intelligence
Image alignment and stitching: a tutorial
Foundations and Trends® in Computer Graphics and Vision
Image Based Quantitative Mosaic Evaluation with Artificial Video
SCIA '09 Proceedings of the 16th Scandinavian Conference on Image Analysis
Three-stage motion deblurring from a video
ACCV'07 Proceedings of the 8th Asian conference on Computer vision - Volume Part II
Color correction using 3d gaussian mixture models
ICIAR'12 Proceedings of the 9th international conference on Image Analysis and Recognition - Volume Part I
| Hi-index | 0.00 |
Inspired by tensor voting, we present luminance voting, a novelapproach for image registration with global and local luminancealignment. The key to our modeless approach is the directestimation of replacement function, by reducing the complexestimation problem to the robust 2D tensor voting in thecorresponding voting spaces. No model for replacement function isassumed. Luminance data are first encoded into 2D ball tensors.Subject to the monotonic constraint only, we vote for an optimalreplacement function by propagating the smoothness constraint usinga dense tensor field. Our method effectively infers missing curvesegments and rejects image outliers without assuming anysimplifying or complex curve model. The voted replacement functionsare used in our iterative registration algorithm for computing thebest warping matrix. Unlike previous approaches, our robust methodcorrects exposure disparity even if the two overlapping images areinitially misaligned. Luminance voting is effective in correctingexposure difference, eliminating vignettes, and thus improvingimage registration. We present results on a variety of images.