Automatic photo orientation detection
View Patent ↗A method determining image orientation includes determining if an image includes an orientation tag and if the tag indicates the image is rotated +90 or −90 degrees. When the image does not include the tag or the tag does not indicate the image is rotated, the method includes determining if a face is detected in the original image and displaying the original image when a face is detected. When a face is not detected, the method includes rotating the image +90 and −90 degrees and detecting a face in the rotated images. When a face is not detected, the method includes applying a classifier to determine the image's proper orientation. When a face is detected in one rotated image, the method includes displaying the rotated image. When a face is detected in both rotated images, the method includes applying the classifier to determine the image's proper orientation.
1. A method for detecting an orientation of a digital image, comprising:
determining, using a processor, if the digital image includes an orientation tag;
when the digital image does not include an orientation tag:
detecting, using the processor, a face in the digital image in its original orientation; and
when a face is detected in the digital image in its original orientation, displaying the digital image in its original orientation without rotation;
when the digital image includes an orientation tag:
determining, using the processor, if the orientation tag indicates the image is rotated to one of (1) +90 degrees and (2) −90 degrees; and
when the orientation tag does not indicate the image is rotated to one of (1) +90 degrees and (2) −90 degrees:
detecting, using the processor, a face in the digital image in its original orientation; and
when a face is detected in the digital image in its original orientation, displaying the digital image in its original orientation without rotation;
when a face is not detected in the digital image in its original orientation:
rotating the digital image +90 degrees to form a first rotated image and detecting, using the processor, a face in the first rotated image;
rotating the digital image −90 degrees to form a second rotated image and detecting, using the processor, a face in the second rotated image;
determining, using the processor, if a face is detected in at least one of the first and the second rotated images;
when a face is not detected in at least one of the first and the second rotated images:
applying a classifier, using the processor, to the digital image in its original orientation to determine probabilities of the digital image being in one of three orientations; and
displaying the digital image in one of the three orientations having the highest probability;
when a face is detected in at least one of the first and the second rotated images:
determining, using the processor, if a face is detected in only one of the first and the second rotated images; and
when a face is detected in only one of the first and the second rotated images, displaying said only one of the first and the second rotated images.
2. The method of claim 1 , further comprising:
when a face is detected in both of the first and the second rotated images:
applying the classifier to the digital image in its original orientation to determine probabilities of the digital image being in one of three orientations; and
displaying the digital image in one of the three orientations having the highest probability.
3. The method of claim 1 , wherein:
the digital image is rotated −90 degrees when the orientation tag indicates a 0 th row is a visual right-hand side of the digital image and a 0 th column is a visual top of the digital image; and
the digital image is rotated +90 degrees when the orientation tag indicates the 0 th row is a visual left-hand side of the digital image and the 0 th column is a visual bottom of the digital image.
4. A non-transitory computer-readable storage medium encoded with executable instructions for execution by a processor to detect an orientation of a digital image, the instructions comprising:
determining if the digital image includes an orientation tag;
when the digital image does not include an orientation tag:
detecting a face in the digital image in its original orientation; and
when a face is detected in the digital image in its original orientation, displaying the digital image in its original orientation without rotation;
when the digital image includes an orientation tag:
determining if the orientation tag indicates the image is rotated to one of (1) +90 degrees and (2) −90 degrees; and
when the orientation tag does not indicate the image is rotated to one of (1) +90 degrees and (2) −90 degrees:
detecting a face in the digital image in its original orientation; and
when a face is detected in the digital image in its original orientation, displaying the digital image in its original orientation without rotation;
when a face is not detected in the digital image in its original orientation:
rotating the digital image +90 degrees to form a first rotated image and detecting a face in the first rotated image;
rotating the digital image −90 degrees to form a second rotated image and detecting a face in the second rotated image;
determining if a face is detected in at least one of the first and the second rotated images;
when a face is not detected in at least one of the first and the second rotated images:
applying a classifier to the digital image in its original orientation to determine probabilities of the digital image being in one of three orientations; and
displaying the digital image in one of the three orientations having the highest probability;
when a face is detected in at least one of the first and the second rotated images:
determining if a face is detected in only one of the first and the second rotated images; and
when a face is detected in only one of the first and the second rotated images, displaying said only one of the first and the second rotated images.
5. The non-transitory computer-readable storage medium of claim 4 , wherein the instructions further comprise:
when a face is detected in both of the first and the second rotated images:
applying the classifier to the digital image in its original orientation to determine probabilities of the digital image being in one of three orientations; and
displaying the digital image in one of the three orientations having the highest probability.
6. The non-transitory computer-readable storage medium of claim 4 , wherein the instructions further comprise:
the digital image is rotated −90 degrees when the orientation tag indicates a 0 th row is a visual right-hand side of the digital image and a 0 th column is a visual top of the digital image; and
the digital image is rotated +90 degrees when the orientation tag indicates the 0 th row is a visual left-hand side of the digital image and the 0 th column is a visual bottom of the digital image.