Video text processing apparatus
Video frames that contain text areas are selected from given video frames by removing redundant frames and non-text frames, the text areas in the selected frames are located by removing false strokes, and text lines in the text areas are extracted and binarized.
1. A text change frame detection apparatus that selects a plurality of video frames including text contents from given video frames, said apparatus comprising:
a first frame removing unit to remove redundant video frames from the given video frames;
a second frame removing unit to remove video frames that do not contain a text area from the given video frames;
a third frame removing unit to detect and remove redundant video frames caused by image shifting from the given video frames; and
an output unit to output remaining video frames as candidate text change frames,
wherein the second frame removing unit includes:
a fast and simple image binarization unit to generate a first binary image of a video frame of the given video frames;
a text line region determination unit to determine a position of a text line region by using a horizontal projection and a vertical projection of the first binary image;
a rebinarization unit to generate a second binary image of every text line region;
a text line confirmation unit to determine validity of a text line region by using a difference between the first binary image and the second binary image and a fill rate of a number of foreground pixels in the text line region to a total number of pixels in the text line region; and
a text frame verification unit to confirm whether a set of continuous video frames are non-text frames that do not contain a text area by using a number of valid text line regions in the set of continuous video frames.
2. A text change frame detection apparatus that selects a plurality of video frames including text contents from given video frames, said apparatus comprising:
a first frame removing unit to remove redundant video frames from the given video frames;
a second frame removing unit to remove video frames that do not contain a text area from the given video frames;
a third frame removing unit to detect and remove redundant video frames caused by image shifting from the given video frames; and
an output unit to output remaining video frames as candidate text change frames,
wherein the third frame removing unit includes:
a fast and simple image binarization unit to generate binary images of two video frames of the given video frames;
a text line vertical position determination unit to determine a vertical position of every text line region by using horizontal projections of the binary images of the two video frames;
a vertical shifting detection unit to determine a vertical offset of image shifting between the two video frames and a similarity of the two video frames in a vertical direction by using correlation between the horizontal projections; and
a horizontal shifting detection unit to determine a horizontal offset of the image shifting and a similarity of the two video frames in a horizontal direction by using correlation between vertical projections of every text line in the binary images of the two video frames,
and the third frame removing unit to remove a similar video frame as a redundant video frame caused by the image shifting.
3. A text change frame detection apparatus that selects a plurality of video frames including text contents from given video frames, said apparatus comprising:
an image block validation unit to calculate a mean value and a variance of a gray level of each of two image blocks in the same position in two video frames of the given video frames, and to determine the two image blocks are a valid block pair that has an ability to show a change of image contents if at least one of two variances of the two image blocks is greater than a first threshold, or if the two variances are smaller than the first threshold and an absolute difference of two mean values of the two image blocks is greater than a second threshold;
an image block similarity measurement unit to calculate a similarity of two image blocks of the valid block pair and to determine whether the two image blocks are similar;
a frame similarity judgment unit to determine whether the two video frames are similar by using a ratio of a number of similar image blocks to a total number of valid block pairs; and
an output unit to output remaining video frames after a similar video frame is removed, as candidate text change frames.
4. A text change frame detection apparatus that selects a plurality of video frames including text contents from given video frames, said apparatus comprising:
a fast and simple image binarization unit to generate a first binary image of a video frame of the given video frames;
a text line region determination unit to determine a position of a text line region by using a horizontal projection and a vertical projection of the first binary image;
a rebinarization unit to generate a second binary image of every text line region;
a text line confirmation unit to determine validity of a text line region by using a difference between the first binary image and the second binary image and a fill rate of a number of foreground pixels in the text line region to a total number of pixels in the text line region;
a text frame verification unit to confirm whether a set of continuous video frames are non-text frames that do not contain a text area by using a number of valid text line regions in the set of continuous video frames; and
an output unit to output remaining video frames after the non-text frames are removed, as candidate text change frames.
5. A text change frame detection apparatus that selects a plurality of video frames including text contents from given video frames, said apparatus comprising:
a fast and simple image binarization unit to generate binary images of two video frames of the given video frames;
a text line vertical position determination unit to determine a vertical position of every text line region by using horizontal projections of the binary images of the two video frames;
a vertical shifting detection unit to determine a vertical offset of image shifting between the two video frames and a similarity of the two video frames in a vertical direction by using correlation between the horizontal projections;
a horizontal shifting detection unit to determine a horizontal offset of the image shifting and a similarity of the two video frames in a horizontal direction by using correlation between vertical projections of every text line in the binary images of the two video frames; and
an output unit to output remaining video frames after a similar video frame is removed, as candidate text change frames.
6. A computer-readable storage medium storing a program used to direct a computer, that selects a plurality of video frames including text contents from given video frames, to perform a process comprising:
removing redundant video frames from the given video frames;
removing video frames that do not contain a text area from the given video frames;
detecting and removing redundant video frames caused by image shifting from the given video frames; and
outputting remaining video frames as candidate text change frames, wherein the removing video frames that do not contain the text area includes:
generating a first binary image of a video frame of the given video frames;
determining a position of a text line region by using a horizontal projection and a vertical projection of the first binary image;
generating a second binary image of every text line region;
determining validity of a text line region by using a difference between the first binary image and the second binary image and a fill rate of a number of foreground pixels in the text line region to a total number of pixels in the text line region; and
confirming whether a set of continuous video frames are non-text frames that do not contain a text area by using a number of valid text line regions in the set of continuous video frames.
7. A computer-readable storage medium storing a program used to direct a computer, that selects a plurality of video frames including text contents from given video frames, to perform a process comprising:
removing redundant video frames from the given video frames;
removing video frames that do not contain a text area from the given video frames;
detecting and removing redundant video frames caused by image shifting from the given video frames; and
outputting remaining video frames as candidate text change frames, wherein the detecting and removing redundant video frames caused by image shifting includes:
generating binary images of two video frames of the given video frames;
determining a vertical position of every text line region by using horizontal projections of the binary images of the two video frames;
determining a vertical offset of image shifting between the two video frames and a similarity of the two video frames in a vertical direction by using correlation between the horizontal projections; and
determining a horizontal offset of the image shifting and a similarity of the two video frames in a horizontal direction by using correlation between vertical projections of every text line in the binary images of the two video frames, and the detecting and removing redundant video frames removes a similar video frame as a redundant video frame caused by the image shifting.
8. A computer-readable storage medium storing a program used to direct a computer, that selects a plurality of video frames including text contents from given video frames, to perform a process comprising:
calculating a mean value and a variance of a gray level of each of two image blocks in the same position in two video frames of the given video frames, and determining the two image blocks are a valid block pair that has an ability to show a change of image contents if at least one of two variances of the two image blocks is greater than a first threshold, or if the two variances are smaller than the first threshold and an absolute difference of two mean values of the two image blocks is greater than a second threshold;
calculating a similarity of two image blocks of the valid block pair and determining whether the two image blocks are similar;
determining whether the two video frames are similar by using a ratio of a number of similar image blocks to a total number of valid block pairs; and
outputting remaining video frames after a similar video frame is removed, as candidate text change frames.
9. A computer-readable storage medium storing a program used to direct a computer, that selects a plurality of video frames including text contents from given video frames, to perform a process comprising:
generating a first binary image of a video frame of the given video frames;
determining a position of a text line region by using a horizontal projection and a vertical projection of the first binary image;
generating a second binary image of every text line region;
determining validity of a text line region by using a difference between the first binary image and the second binary image and a fill rate of a number of foreground pixels in the text line region to a total number of pixels in the text line region;
confirming whether a set of continuous video frames are non-text frames that do not contain a text area by using a number of valid text line regions in the set of continuous video frames; and
outputting remaining video frames after the non-text frames are removed, as candidate text change frames.
10. A computer-readable storage medium storing a program used to direct a computer, that selects a plurality of video frames including text contents from given video frames, to perform a process comprising:
generating binary images of two video frames of the given video frames;
determining a vertical position of every text line region by using horizontal projections of the binary images of the two video frames;
determining a vertical offset of image shifting between the two video frames and a similarity of the two video frames in a vertical direction by using correlation between the horizontal projections;
determining a horizontal offset of the image shifting and a similarity of the two video frames in a horizontal direction by using correlation between vertical projections of every text line in the binary images of the two video frames; and
outputting remaining video frames after a similar video frame is removed, as candidate text change frames.