1. Field of the Invention
The present invention relates to a signal processing of motion-compensated frame rate up-conversion, and more particularly, to an improved signal-processing method for displaying film-originated video signals without motion discontinuities.
2. Description of the Related Art
To convert motion-picture-film-source television signals for display on computer-type monitors and television sets having increased frame rate displays, particularly frame rates of 72 Hz, 96 Hz and 120 Hz, a deinterlacing technique is usually implemented to progressively scan the display. A popular method to deinterlace an interlaced NTSC television signal having 24 frames per second is to merge opposite polarity pairs of interlaced parts coming from the same film source. Each merged pair is repeated at least twice to form a 3-2 pull-down pattern. Thus, the deinterlaced signals have the same frame rate as the original interlaced display signals. Sometimes the deinterlaced 60 Hz progressive scan signals are doubled to 120 Hz frame rate. As a result, a 6-4 pattern is formed. However, a motion discontinuity could be visible in the 6-4 pattern.
The method for motion-compensated frame rate up-conversion in accordance with an embodiment of the present invention comprises the following steps: detecting a film-mode video signal composing a sequence of 3-2 pull-down frames, extracting a sequence of feature frames from the sequence of 3-2 pull-down frames, calculating a motion vector for each pair of the sequence of feature frames, and interpolating a plurality of intermediate frames between each pair of the sequence of feature frames based on the corresponding motion vector.
The method for motion-compensated frame rate up-conversion for a film-oriented video signal in accordance with another embodiment of the present invention comprises the following steps: detecting the film-oriented video signal having a 3-2 pull-down pattern, doubling a frame rate of the film-oriented video signal maintaining the 3-2 pull-down pattern into a 6-4 pattern, wherein the film-oriented video signal comprises a plurality of 6-sequences and 4-sequences arranged alternately, and replacing four consecutive frames covering each boundary of the 6-sequences and the 4-sequences of the film-oriented video signal with four compensated frames to form a compensated video signal.
The invention will be described in accordance with the appended drawings in which:
a) to 1(c) depict the forming process for motion-compensated frame rate up-conversion in accordance with an embodiment of the present invention;
a) to 2(c) depict the detecting method for the film-oriented video signals having a 3-2 pull-down pattern in accordance with one embodiment of the present invention; and
a) and 3(b) show examples of detecting the film-oriented video signal having a 3-2 pull-down pattern.
a) to 1(c) depict the forming process for motion-compensated frame rate up-conversion in accordance with an embodiment of the present invention.
To form the present pattern in accordance with one embodiment of the present invention, a film-mode video signal comprising a sequence of 3-2 pull-down frames is detected and a sequence of feature frames from the sequence of 3-2 pull-down frames is extracted first. Thereafter, a motion vector for each pair of the sequence of feature frames is calculated, and a plurality of intermediate frames between each pair of the sequence of feature frames based on the corresponding motion vector is interpolated. The embodiment of the present invention performs a film-mode detection before the frame rate conversion, and performs actions of motion estimation and compensation to a film source to keep the motion trajectory images continuous. Another embodiment of the present invention forms the pattern by doubling a frame rate of the film-oriented video signal maintaining the 3-2 pull-down pattern into a 6-4 pattern, where the film-oriented video signal comprises a plurality of 6-sequences and 4-sequences arranged alternately. Thereafter, four consecutive frames covering each boundary of the 6-sequences and the 4-sequences of the film-oriented video signal are replaced with four compensated frames. The compensated video signal has an uncompensated frame in each of the 6-sequences and the 4-sequences except for the first sequence and the last sequence thereof.
a) to 2(c) depict the detecting method for the film-oriented video signals having a 3-2 pull-down pattern in accordance with one embodiment of the present invention. The method is mainly made by receiving a source video signal comprising a plurality of frames, calculating a duplication label for each frame of the source video signal, and identifying the source video signal as the film-oriented video signal having a 3-2 sequence. The flow chart in
if|F(x,t−T)−F(x,T)|<FILM_DIFF then (count++)
Where x represents the location of the pixel, T represents the frame interval, and the parameter FILM_DIFF represents an acceptable threshold value. An absolute value of the difference between the gray level values of each pixel of the current frame and the corresponding pixel of the previous frame is obtained. Also, the count flag increases by 1 if the absolute value is smaller than a first threshold FILM_DIFF. The following algorithm illustrates that if most pixels in these two frames are equal, then these two frames are regarded as duplicates.
Where the parameter to1_count represents the total number of a frame, and FILM_THD represents an acceptable threshold value. First of all, whether or not the current frame and the previous frame are identified as two duplicate frames is in accordance with the ratio of the count flag to the total pixel number of the current frame. The duplication label in accordance with the ratio of the count flag is adjusted to the total pixel number of the current frame. The above steps are repeated for each frame of the source video signals.
The flow chart in
Where the parameter film_dcount represents the number of continuous duplicates, film_pattern (2 or 3) represents a fixed pattern, and film_tallycount represents the number of matching the patterns.
The flow chart in
Where the parameter FILM_TCOUNT_H represents a high-level threshold, and FILM_TCOUNT_L represents a low-level threshold.
a) and 3(b) show examples of detecting the film-oriented video signal having a 3-2 pull-down pattern. Those examples are demonstrated based on the algorithms above and the flow charts of
The above-described embodiments of the present invention are intended to be illustrative only. Numerous alternative embodiments may be devised by persons skilled in the art without departing from the scope of the following claims.