Methods for slide image classification
View Patent ↗A method of classifying images as slide images or non-slide images can be done by capturing a video stream, analyzing an image of the video stream and determining whether the image is a slide image. The analyzing can be based on text height and/or edge region(s) of the image.
1. A device comprising:
a processor to analyze images of a video stream, wherein a height of text located in at least one predetermined edge region of the video stream is analyzed in the images to classify the images of the video stream as being slide images or non-slide images, wherein the edge region comprises the outermost north, east, west or south borders of the video stream; and
a memory to store the slide images.
2. The device of claim 1 , wherein the processor performs optical character recognition on the images to return bounded boxes identifying the text in the images.
3. The device of claim 1 , wherein the processor compares an average height of text in an image of the images to a threshold.
4. The device of claim 3 , wherein the processor determines images whose average text size is greater than the threshold to be slide images.
5. The device of claim 1 , wherein the processor samples the video stream to obtain the images.
6. The device of claim 1 , wherein the processor creates histograms indicating the bounding box sizes of the text of the images and does a nearest neighbor classification on the histograms to determine whether the images are slide images.
7. The device of claim 1 , wherein the at least one predetermined edge region of the video stream includes a predetermined number of rows or columns.
8. The device of claim 1 , wherein a plurality of predetermined edge regions are analyzed.
9. The device of claim 1 , wherein luminance values of the at least one predetermined edge region of the video stream are examined.
10. The device of claim 1 , wherein the processor samples the video stream to obtain the images.
11. The device of claim 1 , wherein the processor uses text height in any region of the video stream to determine a first set of slide images, and uses characteristics of at least one predetermined edge region around the outer border of the first set of slide images to evaluate remaining images to classify additional slide images.
12. A computer-readable medium having a program of instructions stored thereon that performs a function, the function comprising the steps of: capturing an input video stream; and
analyzing a height of text located in at least one predetermined edge region of the video stream to classify the images of the video stream as being slide images or non-slide images, wherein the edge region comprises the outermost north, east, west or south borders of the video stream.
13. The program of claim 12 , wherein the average height of text in the image is compared to a threshold.
14. A method for classifying images of a video stream as being slide images and non-slide images, the method comprising:
capturing an input video stream; and analyzing a height of text located in at least one predetermined edge region of the video stream to classify the images of the video stream as being slide images or non-slide images, wherein the edge region comprises the outermost north, east, west or south borders of the video stream.
15. The method of claim 14 , further comprising:
performing optical character recognition on the images to return bounded boxes identifying the text in the images.
16. The method of claim 14 , further comprising:
comparing an average height of text in an image of the images to a threshold.
17. The method of claim 16 , further comprising:
determining images whose average text size is greater than the threshold to be slide images.
18. The method of claim 14 , further comprising:
sampling the video stream to obtain the images.
19. The method of claim 14 , further comprising:
creating histograms indicating the bounding box sizes of the text of the images; and
performing a nearest neighbor classification on the histograms to determine whether the images are slide images.
20. The method of claim 14 , wherein the at least one predetermined edge region of the video steam includes a predetermined number of rows or columns.
21. The method of claim 14 , wherein a plurality of predetermined edge regions are analyzed.
22. The method of claim 14 , further comprising:
examining the luminance values of the at least one predetermined edge region of the video stream.