Television – Format conversion – Line doublers type
Reexamination Certificate
1998-03-31
2002-01-22
Eisenzopf, Reinhard J. (Department: 2614)
Television
Format conversion
Line doublers type
C348S449000, C348S451000, C348S452000
Reexamination Certificate
active
06340990
ABSTRACT:
FIELD OF INVENTION
The present invention relates to a method for converting an interlaced television display to a progressive, scan, or non-interlaced display where artifacts are removed if the source of the signal is from a movie film to television converter, a video recording of a movie, a live camera, or a camera output captured on a video recording system, or any sequence of the above.
BACKGROUND OF THE INVENTION
The NTSC system for television broadcasting of images or frames that contain 525 lines involves transmitting a first field of 262.5 rows of even parity followed by a second field of 262.5 rows of odd parity. The fields are transmitted at a rate of 60 Hz, while the frames are transmitted at a rate of 30 Hz. On small television displays, the eye can easily integrate the sequence of lower resolution odd and even row images to give the impression of higher resolution 525-row images. However on large displays, the eye can see the effects of interlace as annoying artifacts because of the larger pixel size. The persistence of the illumination of the image on larger screens is shorter and leads to a noticeable flicker.
It is desirable on large screen displays to deinterlace the image by merging odd parity and even parity fields together to form fields with the full resolution. However, it is well known that if the merging is done simply by interline averaging or simply overlaying the odd and even fields, there is a severe degradation in the picture quality because of the motion of objects from one field to the next.
The problem of deinterlacing is to construct the missing even lines from odd fields, and the missing odd lines from the even fields. A number of systems have been disclosed that attempt to solve the problems of deinterlacing. One method such as disclosed in U.S. Pat. No. 4,057,835 involves a motion detector that provides a signal for those areas that contain a moving object. Those areas are processed differently than an area where there is no motion. For example, in an area of an odd field where there is no motion, the missing lines could simply be computed by averaging each corresponding line in the two adjacent even fields. Where there is motion, the missing lines can be computed from the lines above and below within the same odd field.
If there is in fact no motion in a region of a sequence of fields, it is always easily possible to reliably detect that case and supply an accurate computation of the missing lines. If there is in fact motion in a region of a field, it is not always possible to detect that motion. Even when the motion is detected, it is not always possible to supply an accurate computation of the missing lines. Thus, errors in deinterlacing will occur in areas where there is motion.
In some systems, motion is detected by analyzing two adjacent fields of the same parity on either side of the field for which the deinterlaced line is undergoing computation. If these two fields are substantially the same, then no motion is inferred. This method works for slowly moving objects. When the object is moving very fast across a still background, the object may be in one field at a certain location, but may not be present at all in the two opposite parity fields immediately before and after the field. The motion detector will sense the same background in both adjacent fields and infer that there is no motion and will erroneously merge the data in the adjacent fields with the field undergoing computation. Artifacts will result as disclosed by Powers in U.S. Pat. No. 4,400,719.
A temporal median filter disclosed by Faroudja in U.S. Pat. No. 4,989,090 also exhibits artifacts when the motion of an object is very fast. The median operates between a first field of one parity a second preceding adjacent field and third subsequent adjacent field, both of opposite parity. For a very fast moving object, over a constant background, the two adjacent fields will have substantially the same pixel values over a neighborhood. Thus the median filter will choose the value of one of the adjacent fields and use it incorrectly as the interlaced value for the first field.
If the source of the video signal is derived from a movie film to television converter, the motion detection problem is much different. A film sequence is transferred at a rate of 24 frames per second. The most common film transmission technique is to use what is called a three to two pull-down ratio where a first film frame is projected for three sequential television fields, and the next film frame is projected for two sequential television fields. The deinterlacing concept of using a motion detector and motion dependent field merging is not the best procedure for film sequences.
One way of handling deinterlacing of film sequences is to first detect that the fields are transmitted in the three to two pull-down ratio by a technique such as disclosed by Faroudja in U.S. Pat. No. 4,982,280. There is no notion between some pairs of fields because the film is guaranteed to be stationary for at least two fields. Thus the next step for deinterlacing film is to associate the corresponding odd and even fields that actually belong to the same film frame and then interleave lines of those fields to form a progressive scan as disclosed by Faroudja in U.S. Pat. No. 4,876,596.
Although Faroudja's method often gives excellent results for film sequences using the most popular three to two pull-down ratio, the method has some drawbacks. Faroudja's method does not work for older Interlaced Telecine film converters that capture 2½ film frames for each pair of television fields. Every third field will contain one film frame in the upper half and the next film frame in the lower half. There is no correct pairing of fields that will eliminate annoying artifacts for these split fields.
Another drawback of Faroudja's method is that it requires a sequence of several television fields before the phase of the fields can be determined with respect to the film frame. When film that had been converted to a video tape recording is edited or when film clippings are interspersed between live video, the phase of the fields can suddenly alter. Thus there could be frequent interruptions of the phase, and the attendant failings of the deinterlacing method could cause annoying artifacts while the phase is being resynchronized.
SUMMARY OF THE INVENTION
The general object of this invention is to overcome some of the drawbacks in the prior art of television display deinterlacing.
Another object of this invention is to provide a motion detector that functions when the video sequence is so rapidly changing that motion cannot be inferred from two adjacent fields of the same parity.
Another object of this invention is to provide one single method that will minimize motion artifacts in both camera video, and video recordings of film in any of the film-to-television converters.
Another object of this invention is to provide a fast detection of movie film so that artifacts are Minimized during switchover when video recordings derived from film are edited or merged with camera video output.
The method of the present invention can be realized in a low cost fine grained programmable massively parallel processor (MPP) such as disclosed by Wilson in U.S. Pat. No. 5,557,734 incorporated herein by reference.
In the current invention, the missing interlaced lines of a field are first supplied by a simple interpolation between adjacent lines of that field. Next, a sequence of steps in the processor forms a first motion detector,that detects areas of motion between fields and substitutes an appropriate correction to those areas in the field that have no motion. Then, a second more complex motion detector that spans a range of five fields is used to test for motion over a longer time period. This motion detector may reverse the decision of the first motion detector. Often when an object is very quickly moving over a still background, neither of the first two motion detectors will detect motion because they erroneously determine that since th
Applied Intelligent Systems Inc.
Eisenzopf Reinhard J.
Lo Linus H.
LandOfFree
System for deinterlacing television signals from camera... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System for deinterlacing television signals from camera..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for deinterlacing television signals from camera... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2832701