Television – Format conversion – Line doublers type
Reexamination Certificate
2001-05-31
2002-10-01
Lee, Michael H. (Department: 2614)
Television
Format conversion
Line doublers type
C348S458000
Reexamination Certificate
active
06459454
ABSTRACT:
BACKGROUND OF THE INVENTION
1. The Field of the Invention
The present invention relates to the field of digital video. More specifically, the present invention relates to systems for adaptively converting interlaced fields of video into progressive frames on a per pixel basis.
2. The Related Art
Video information may be represented by progressive video or interlaced video. Modern computer monitors typically display progressive video. Conventional television monitors and older computer monitors typically display interlaced video. High definition television may display both interlaced and progressive video.
Progressive video includes a series of frames, where each frame is drawn as consecutive lines from top to bottom. In interlaced video, each frame is divided into a number of fields. Typically, the frame is divided into two fields, one field containing half of the lines (e.g., the even numbered lines), and the other field containing the other half of the lines (e.g., the odd numbered lines). The interlaced video, however, is still temporally ordered so that neighboring interlaced fields may represent video information sampled at different times.
There is often a need to convert interlaced video into progressive video and vice versa. For example, suppose a television broadcaster transmits a conventional television program as a series of interlaced fields. If these interlaced fields are to be displayed on a modem computer monitor (or on a high definition television display) that displays progressive frames, the interlaced fields must be converted into progressive frames.
The conversion involves using one or more fields of interlaced video to generate a frame of progressive video and repeating the process so that a stream of interlaced video is converted into a stream of progressive video. This conversion is often called “deinterlacing”. There are several conventional methods of deinterlacing.
One conventional deinterlacing method is called “scan line interpolation” in which the lines of a single interlaced field are duplicated to form a first half of the lines in the progressive frame. The second half of the lines in the progressive frame are formed by simply duplicating the same field again and inserting the field offset by one line into the second half of the lines to complete the progressive frame. This basic form of scan line interpolation is computationally straightforward and thus uses little, if any, processor resources. However, the vertical resolution of the progressive frame is only half of what the display is capable of displaying.
One variation on the scan line interpolation method is that the second half of the lines in the progressive frame are generated by interpolating (e.g., averaging) the neighboring lines in the interlaced field. This requires somewhat more computational resources, but results in a relatively smooth image. Still, the vertical resolution is only half of what the display is capable of displaying.
One deinterlacing method that improves vertical resolution over scan line interpolation is called “field line merging” in which lines from two consecutive fields are interweaved to form a progressive frame. However, the video information in the first field is not sampled at the exact same moment as the video information in the second field. If there is little movement in the image between the first and second fields, then field line merging tends to produce a quality image at relatively little processing costs. On the other hand, if there is movement between the first and second fields, simply combining fields will not result in a high fidelity progressive frame since half the lines in the frame represent the video data at a given time, and half the lines in the frame represent a significantly different state at a different time.
Higher processing methods use complex motion compensation algorithms to determine where in the image there is motion, and where there is not. For those areas where there is no motion, field line merging is used because of its improved vertical resolution. For those areas where there is motion, scan line interpolation is used since it eliminates the motion artifacts that would be caused by field line merging. Such motion compensation algorithms may be implemented by the motion estimation block of an MPEG encoder. However, such complex motion compensation methods require large amounts of processing and memory resources.
Therefore, what are desired are systems for deinterlacing to provide a relatively high fidelity progressive frame without having to dedicate the processor and memory resources required by complex motion compensation algorithms.
SUMMARY OF THE INVENTION
The principles of the present invention provide for the adaptive deinterlacing of interlaced video to generate a progressive frame on a per pixel basis. In a first embodiment of the present invention, two consecutive fields of interlaced video are converted into a frame of progressive video. One of the fields is replicated to generate half the lines in the progressive frame. Each of the pixels in the other half of the progressive frame are generated pixel-by-pixel.
Specifically, for a given output position of the pixel in the other half of the progressive frame, a correlation is estimated between the corresponding pixel in the non-replicated field and at least one vertically adjacent pixel of the replicated field, and optionally one or more vertically adjacent pixels in the non-replicated fields. In one example, a window of pixels one pixel wide by five pixels high is evaluated centering on the pixel in the non-replicated field that corresponds to the output pixel position.
A value is then assigned to the output pixel that corresponds to the output position, the value depending on the correlation. The deinterlacing in accordance with the present invention interpolates between scan line interpolation and field merging depending on the correlation. For example, if there is a high vertical correlation, then more of field merging is performed for that pixel since a high correlation suggests less likelihood of movement at that pixel position. If there is a low vertical correlation, then more of scan line interpolation is performed for that pixel since a low correlation suggests more likelihood of movement at that pixel position. If there is moderate correlation, a balance of scan line interpolation and field merging is performed. This process is repeated for each pixel in the other half of the progressive frame until the entire progressive frame is generated.
Thus, unlike pure scan line interpolation or pure field line merging, the deinterlacing in accordance with the present invention adaptively uses a portion of each method depending on how much motion is detected at the pixel. The mechanism for estimating motion in accordance with the present invention is not as sophisticated as the conventional complex motion compensation methods. However, the mechanism for estimating motion compensation in accordance with the present invention provides suitable motion estimation for many video applications. In addition, the deinterlacing algorithm in accordance with the present invention does not require the extensive processing and memory resources that the complex motion compensation methods require. Therefore, the deinterlacing of the present invention is ideally suited for video applications in which processing and memory resources are limited.
In a second embodiment of the invention, three consecutive input fields of interlaced video are converted into two output fields of interlaced video. The second temporal input field is replicated to produce a first of the two output fields. The other field is generated on a per pixel basis.
Specifically, for a given output pixel corresponding to an output position of the second output field, at least one pixel of the second temporal input field that is vertically adjacent to the output position of the second output field is used to determine which of the first temporal input field and third temporal input field more closely correlates to
Lee Michael H.
WebTV Networks Inc.
Workman & Nydegger & Seeley
LandOfFree
Systems for adaptively deinterlacing video on a per pixel basis does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Systems for adaptively deinterlacing video on a per pixel basis, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Systems for adaptively deinterlacing video on a per pixel basis will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2941343