Systems and methods for illuminant-invariant model estimation
A method is described. The method includes determining normalized radiance of an image sequence based on a camera response function (CRF). The method also includes determining one or more reliability images of the image sequence based on a reliability function corresponding to the CRF. The method further includes extracting features based on the normalized radiance of the image sequence. The method additionally includes optimizing a model based on the extracted features and the reliability images.
1. A method, comprising:
determining normalized radiance of an image sequence based on a camera response function (CRF);
determining one or more reliability images of the image sequence based on a reliability function corresponding to the CRF;
extracting features based on the normalized radiance of the image sequence; and
optimizing a model based on the extracted features and the reliability images.
2. The method of claim 1 , wherein optimizing the model based on the extracted features and the reliability images comprises estimating a camera pose based on the extracted features and the reliability images.
3. The method of claim 2 , wherein estimating the camera pose comprises minimizing an objective function that includes the extracted features, the reliability images and a point-to-plane distance with respect to a transformation of point clouds associated with the image sequence.
4. The method of claim 3 , wherein the objective function to estimate the camera pose is formulated as a least square system.
5. The method of claim 1 , wherein determining normalized radiance of the image sequence based on the CRF comprises:
determining a first normalized radiance of a source image in the image sequence based on the CRF; and
determining a second normalized radiance of a target image in the image sequence based on the CRF.
6. The method of claim 5 , wherein determining reliability images of the image sequence based on the reliability function corresponding to the CRF comprises:
determining a first reliability image of the source image in the image sequence based on the reliability function; and
determining a second reliability image of the target image in the image sequence based on the reliability function.
7. The method of claim 5 , wherein extracting features based on the normalized radiance of the image sequence comprises:
extracting features from the first normalized radiance; and
extracting features from the second normalized radiance.
8. The method of claim 5 , wherein optimizing the model is based on features extracted from the first normalized radiance and the second normalized radiance, a first reliability image associated with the first normalized radiance and a second reliability image associated with the second normalized radiance.
9. The method of claim 1 , wherein optimizing the model is further based on depth data from the image sequence.
10. The method of claim 1 , wherein the optimal model comprises one of feature matching, homography estimation, image registration, object segmentation or object detection.
11. An electronic device, comprising:
a memory; and
a processor in communication with the memory, the processor configured to:
determine normalized radiance of an image sequence based on a camera response function (CRF);
determine one or more reliability images of the image sequence based on a reliability function corresponding to the CRF;
extract features based on the normalized radiance of the image sequence; and
optimize a model based on the extracted features and the reliability images.
12. The electronic device of claim 11 , wherein optimizing the model based on the extracted features and the reliability images comprises estimating a camera pose based on the extracted features and the reliability images.
13. The electronic device of claim 11 , wherein determining normalized radiance of the image sequence comprises:
determining a first normalized radiance of a source image in the image sequence based on the CRF; and
determining a second normalized radiance of a target image in the image sequence based on the CRF.
14. The electronic device of claim 13 , wherein optimizing the model is based on features extracted from the first normalized radiance and the second normalized radiance, a first reliability image associated with the first normalized radiance and a second reliability image associated with the second normalized radiance.
15. The electronic device of claim 11 , wherein optimizing the model is further based on depth data from the image sequence.
16. The electronic device of claim 11 , wherein the optimal model comprises one of feature matching, homography estimation, image registration, object segmentation or object detection.