WO1995027374A1

WO1995027374A1 - Video image sampling lattice for vhs format magnetic tape applications

Info

Publication number: WO1995027374A1
Application number: PCT/US1995/003808
Authority: WO
Inventors: Henry N. Kannapell
Original assignee: Apple Computer, Inc.
Priority date: 1994-03-30
Filing date: 1995-03-27
Publication date: 1995-10-12
Also published as: AU2197495A; JPH08511410A; EP0701763A1

Abstract

A sampling lattice for describing a video image presumed to be destined for VHS format magnetic tape recording. The lattice is interleaved and may run at about fifty or sixty fields per second. The entire signal may be rendered at about 4.5 megasamples per second.

Description

VIDEO IMAGE SAMPLING LATTICE FOR VHS FORMAT
MAGNETIC TAPE APPLICATIONS
Background of the Invention
The present invention relates generally to sampling of video images, and more particularly to sampling of video images for magnetic tape recording applications.

As is well known, video images are represented electronically using a temporal succession of frames. Each frame consists of a plurality of continuous horizontal lines. A video image may be sampled temporally and vertically. Such sampling is necessary in order to represent an image having continuous extent in three dimensions (horizontal, vertical and time), using signals that are continuous in only one dimension (time). For each point on a horizontal line, three signal values are necessary to render a color image. The brightness of the point may be rendered using one of the signals, termed luminance, and the color information by two signals termed chrominance.

In the National Television System Committee (NTSC) standard, the video signal runs at 29.97 frames per second. In other countries, a frequency of 25 frames per second is used, which does not preclude the use of the present invention. A field, which is one-half of a frame, is presented at each such interval. One field contains odd numbered lines of the frame, while the other field of the same frame contains the even numbered lines. The fields are interleaved.

The position and time of the lines are shown in Figure 1 as an array of points in a plot having time and vertical position as the coordinate axes. The coordinates of each point are the time and vertical position of a horizontal line of the video signal. For example, there are lines at position 1 at times 1/60 seconds (sec) and 3/60 sec, and there are lines at position 2 at times 2/60 sec and 4/60 sec. Such an array will be referred to as a sampling lattice.

VHS video tape format recording has limited bandwidth for the video signals. The bandwidth is around 1/2 to 1/3 the bandwidth of a broadcast video signal, and is an order of magnitude lower than the bandwidth of a 14"
Macintosh or IBM PC computer monitor, which is about 27 megabytes per second (Mbytes/sec). For example, the VHS tape recording luminance bandwidth is only about 2 megahertz (MHz).

Sampling theory indicates that at a sampling rate of only about 4 megasamples per second (4 Msamples/sec), the full bandwidth of the VHS format video signal may be described. Sampling theory is discussed in C.

Shannon,"A mathematical theory of communication,"Bell Svstem Technical
Journal, vol. 27. pp. 379. 623. 1948. Assigning half of the signal energy to luminance, and half to chrominance, which is part of prior art, the two chroma components are rendered at half the luminance bandwidth, which may be 2.25
Msamples/sec. These rates result in 240 sampling points per line in the horizontal direction for luminance, and 120 sampling points per line for each of the chrominance signals. The vertical line rate is fixed by video standards to be 487 active lines of video (486 lines and two half lines) for NTSC video. This gives a higher resolution vertically than horizontally.

In order to take advantage of this difference, render the image resolution isotropically, and to lower memory and disk usage and processor time requirements for digital video processing, it is common practice to discard one field from each frame. This sampling method will be referred to as alternate field sampling. The corresponding sampling lattice is shown in Figure 2 using solid dots whose coordinates are the positions and the times of the lines, and x's indicate the positions and times of the lines that have been discarded. For example, the lines at odd positions and times that are odd multiples of 1/60 sec are retained, and lines at even positions and times that are even multiples of 1/60 sec are discarded. The temporal sampling rate of this lattice, and the highest frequency that can be rendered without aliasing, are one-half those of the NTSC standard sampling lattice.

Aliasing occurs when the frequency being sampled is greater than one half the sampling frequency. In such a case, the frequency being sampled and a lower frequency cannot be distinguished from each other after sampling.

Accordingly, an object of the present invention is to provide a method for sampling video images having the same vertical resolution and horizontal resolution as alternate field sampling and twice the temporal sampling rate and resolution of alternate field sampling, while maintaining the same overall sampling rate.

Another object of the present invention is to provide a method for sampling video images that provides flexibility in the choice of vertical resolution.

Additional objects and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objects and advantages of the invention may be realized and obtained by means of the instrumentalities and combinations particularly pointed out in the claims.

Summary of the Invention
The present invention is a method of sampling a video image, comprising sampling the image according to a sampling lattice running at a frame rate appropriate for magnetic tape recording. Each frame of the sampling lattice comprises two interleaved fields having uniformly spaced horizontal lines. The lines of one of the fields coincide with a subset of a set of horizontal lines of a corresponding field appropriate for the magnetic tape recording.

Brief Description of the Drawings
The accompanying drawings, which are incorporated in and constitute a part of the specification, schematically illustrate a preferred embodiment of the invention and, together with the general description given above and the detailed description of the preferred embodiment given below, serve to explain the principles of the invention.

Figure 1 is a diagram of a standard NTSC vertical and temporal sampling lattice.

Figure 2 is a diagram of the vertical and temporal alternate field sampling lattice superimposed onto a standard NTSC sampling lattice.

Figure 3 is a diagram of a vertical and temporal sampling lattice according to the present invention superimposed onto a standard NTSC sampling lattice.

Figure 4 is the Fourier transform of the NTSC sampling lattice of Figure 1.

Figure 5 is a diagram of the vertical and temporal frequency content of a continuous time varying image.

Figure 6 is a diagram of the vertical and temporal frequency content of the image of Figure 5 after being sampled with the NTSC sampling lattice of
Figure 1.

Figure 7 illustrates the truncation of the spectrum of the signal of Figure 5 by low-pass filtering.

Figure 8 is a diagram of the fourier transform of the sampling lattice of
Figure 2.

Figures 9A-9G are a succession of views of a rotating wheel.

Figures 10A-10D are a succession of views of the wheel of Figures 9A-9G at one half the sampling rate of Figures 9A-9G.

Figure 11 is a diagram of the fourier transform of the sampling lattice according to the present invention of Figure 3.

Figure 12 is a schematic block diagram of the conversion of an NTSC television signal to a digital signal sampled using the sampling lattice of the present invention.

Figures 13A-13H illustrate intermediate steps in the conversion of Figure 12.

Figure 14 is a schematic block diagram of the conversion to an NTSC television signal of a digital signal sampled using the sampling lattice of the present invention.

Figures 15A-15G illustrate intermediate steps in the conversion of Figure 14.

Description of the Preferred Embodiments
The present invention will be described in terms of a preferred embodiment. The preferred embodiment is a method for sampling video images using an interleaved sampling lattice with reduced horizontal line density. Such a sampling lattice is shown in Figure 3.

The positions of the lines of the lattice of Figure 3 are indicated by solid and hollow dots. Each frame contains two fields. One of the two fields is formed by samples shown as solid dots while the other field of the frame is formed by samples shown as hollow dots which are not coincident with the original samples. The vertical spacing of the horizontal lines is higher than the vertical spacing of the lines of the standard lattice by an integer factor such as, in the present case, two. This is true whenever the vertical line spacing is higher than the vertical spacing of the lines of the standard lattice by an even factor. For example, in Figure 3, the line spacing is twice the standard line spacing, and the solid dots and the x's form a standard sampling lattice. The x's indicate the location and the time of those lines of the standard lattice that have been discarded.

The solid dots are every other line of every other field of the standard sampling lattice such as, for example, the odd fields. The hollow dots are at the centers of the rectangles formed by the solid dots, and do not correspond to lines of the standard lattice.

For example, there are lines at positions 1,5 and 9 and times, such as 1/60 sec, 3/60 sec, 5/60 sec, 7/60 sec, that are odd multiples of 1/60 sec. These lines coincide with lines of a standard sampling lattice, of which the lines indicated by x's have been discarded. The lines at positions 3,7 and 11 and times, such as 2/60 sec, 4/60 sec, 6/60 sec, that are even multiples of 1/60 sec do not correspond to lines in the standard sampling lattice. The hollow dot corresponding to the line at position 3 and time 2/60 sec is at the center of the rectangle formed by the solid dots at positions 1 and 5, and times 1/60 sec and 3/60 sec.

In other embodiments of the present invention, in which the vertical line spacing is higher than the vertical spacing of the lines of the standard lattice by an odd factor, all of the samples of the sampling lattice coincide with samples from a standard sampling lattice. It is possible to convert a signal sampled using the standard lattice to the sampling lattice of the present invention by suitable low-pass filtering of the image. Prior to recording, the signal sampled according to the present invention may be converted to the standard sampling lattice. Such filtering is well known to those skilled in the art.

To better understand the advantages of the sampling lattice of the present invention, it is helpful to examine the effects of the various kinds of sampling lattices in the frequency domain by taking their Fourier transforms.

The sampling of a continuous time dependent image may be regarded as the multiplication of that image by a three-dimensional sampling signal having horizontal, vertical and temporal extent, and that is constant on the horizontal lines of the sampling lattice and zero everywhere else. This sampling signal is similar to a sampled video image but is uniform. In the spatial and temporal frequency domain, the multiplication of the image with the sampling signal becomes a convolution of the Fourier transforms of the image and of the sampling signal. Figure 4 illustrates the non-zero vertical and temporal frequency components of the sampling lattice of Figure 1. In Figures 4 and 5, the horizontal axis corresponds to the temporal frequency in Hertz, and the vertical axis corresponds to periodicity in cycles per line spacing (line¯1).

The non-zero vertical and temporal frequency components of the sampling lattice of
Figure 1 are discrete, so that Figure 4 is an array of points. An arbitrary example of a possible frequency content of an image, to be used in the discussion that follows, is shown in Figure 5. The convolution of the signals of Figures 4 and 5 is shown in Figure 6. It can be seen that a copy of the spectrum of Figure 5 is centered on each point of the frequency domain lattice of Figure 4. If the image has sufficiently high frequency components, aliasing may occur. As shown in
Figure 6, aliasing is due to an overlap between spectra of the image of Figure 5, centered on different points of the frequency domain sampling lattice of Figure 4.

In the absence of aliasing, the original signal of Figure 5 could have been recovered from the sampled signal of Figure 6 by low-pass filtering. If aliasing has occurred, recovery of the original signal is not possible in general. Examples of aliasing are moire fringes and the appearance that a rotating spoked wheel is stationary or moving at an incorrect angular velocity when its image is sampled in time.

In order to avoid aliasing, the original signal may be low-pass filtered before sampling, at the price of losing the high frequency information. For applications involving signals with real values, the pass band of the filter should have inversion symmetry about the origin. The pass band of a filter that prevents aliasing in the case of the standard sampling is the cross-hatched area in Figure 4. Its effect on the spectrum of Figure 5 is to truncate it, as shown in
Figure 7.

As discussed, in alternate field sampling, one of the two fields of every frame is discarded, resulting in the temporal and vertical sampling lattice of
Figure 2. In Figure 2, the solid dots indicate lines that are retained, and the x's indicate lines that were discarded. The Fourier transform of the alternate field sampling lattice is shown in Figure 8. The range of frequencies that can be sampled without aliasing is the cross-hatched square centered at the origin. The area of this square is one-half of the area of the cross-hatched diamond in Figure 4, because the sampling lattice point density is reduced by one half. The maximum frequencies both in time and in the vertical direction that can be rendered without aliasing are one half of the corresponding maximum frequencies in Figure 4.

It follows that images having temporal frequency content between 15 and 30 Hz are aliased when sampled with the alternate field sampling lattice but not when sampled using the standard sampling lattice which, however, requires twice as much bandwidth. As will be shown below, such images are also not aliased when sampled according to the present invention, using the same bandwidth as the alternate field sampling lattice.

Figures 9A-9G and 10A-lOD illustrate aliasing introduced by the reduction of the temporal frequency from the 60 Hz of standard sampling to the 30 Hz of alternate field sampling. Figures 9A-9G are a succession of views of a spoked wheel rotating clockwise. The views could have been taken, for example, at 1/60 second intervals. Had the wheel been imaged at a rate of 30 views per second, the views of Figure 10A-10D would have resulted. The wheel is the same, Figure 10A is identical to Figure 9A, Figure 10B is identical to Figure 9C,
Figure 10C is identical to Figure 9E, and Figure 10D is identical to Figure 9G. In the views of Figures 10A-lOD, the wheel appears to be moving counterclockwise and at a lower angular speed and thus is rendered incorrectly.

Figure 11 shows the fourier transform of the sampling lattice according to the present invention that is shown using solid and hollow dots in Figure 3. The shaded diamond at the origin gives the passband for the low-pass filtering necessary to prevent aliasing. As in the case of alternate field sampling, the resulting signal has 1/2 the bandwidth of the standard signal, and for purposes of
VHS videotape recording, its vertical resolution is closer to the horizontal resolution. For stationary objects, the vertical resolution of this sampling lattice is the same as that of the alternate frame sampling lattice. The maximum frequency that can be rendered without aliasing is twice that of the standard samplinglattice.

The lines of the sampling lattice of the present invention may be continuous. As discussed above, the lines may also be sampled in the horizontal direction, preferably using about 240 sampling points per line for luminance, and about 120 sampling points per line for each of the chrominance signals. Thus each sampling point would contribute to the luminance signal and half the sampling points would contribute to each chrominance signal. The same sampling points preferably contribute to both chrominance signals.

The entire signal can be rendered at about 4.5 megasamples/second (MSamples/sec). This data rate is extremely significant, as it is the data rate that a current high-end consumer disk drive can provide on average.

The sampling lattice of the present invention may be implemented by means of a concatenation of three signal processing elements. By way of example, Figure 12 is a schematic block diagram for the process of taking a standard digitized television signal 20 and creating a reduced bandwidth signal 22 according to the present invention. The process is composed of three blocks.

The first block 24 is a zero stuff. Original signal 20 is interleaved. In order to create full frames for each field time, zeros are stuffed into the even lines 30 for the first field 32 and into the odd lines 34 of the second field 36, as shown in
Figure 13A. The successive fields are separated by a duration T of about 1/60 sec, have a width W that may be 720 horizontal pixels to retain the full bandwidth of NTSC signals and a height H of 480 lines.

Next, since the signal will be sampled at a lower bandwidth, the signal is low-pass filtered prior to sampling step 28, as shown by block 26 of Figure 12.

The signal will be resampled in the three dimensions of horizontal, vertical and time, so a three-dimensional low-pass filtering is needed. The three-dimensional' filter may be separable in which case it may be synthesized using three onedimensional filters, as shown in Figures 13B-13D, which graphically illustrate the filter windows with which the signal is convolved and the resulting horizontal (Hh), vertical (Hv) and temporal (Ht) transfer functions of the filters. As shown, the horizontal and vertical filters may have seven taps, and the temporal filter may have five taps. The gain of the vertical filter compensates for the bit stuffing previously done. As shown, the horizontal and vertical filters may be implemented using seven taps, and the temporal filter using five taps. The result is a low-pass filtered signal, illustrated in Figure 13E.

The sampling of block 28 of Figure 12 is done in two parts. The horizontal sampling lattice is independent and is done first, as shown in Figure 13F. Then the time and vertical sampling, which are interrelated, are done together as shown in Figure 13G. The output reduced bandwidth signal 22 is shown in
Figure 13H.

The television signal may be reconstructed as shown in Figure 14, using a method similar to that of Figure 12. Reduced signal 22 is first zero stuffed to full size, as shown by block 38. This means the 240 pixel horizontal lines are zero stuffed to 720 pixel lines, and the vertical and time dimensions are zero stuffed to 480 lines by 60 fields per second. The result is the signal 44 of Figure 15A.

Signal 44 is passed through a three-dimensional low-pass filter that can be separated into three one-dimensional filters, as shown by block 40 of Figure 14. The three one-dimensional filters have gain appropriate to the amount of zero stuffing that has taken place. Their transfer functions are shown in
Figures 15B-15D. For example, the gain of the horizontal filter must be three, as shown in Figure 15B, since there are three pixels created for every one in the original image. It is also possible to implement the time and vertical filtering simultaneously, by convolving these two filters together.

The resulting low-pass signal 46 (Figures 14 and 15E) is then resampled with a standard television sampling lattice shown in Figure 15F, to obtain a reconstructed signal 48 (Figures 14 and 15G) which is similar to input signal 20 of Figure 12 but with a bandwidth limited by the bandwidth of VHS recording.

It is possible to combine several of the above steps to reduce the amount of data traffic that will occur in the system without departing from the scope of the present invention. The design of the filters is known by those skilled in the art and may also be found in standard textbooks on digital filter design such as
Maurice Bellanger, Digital Processing of Signals-Theory and Practice, John
Wiley and Sons, New York, 1984.

In summary, a method for nearly optimal sampling of video images destined for videotape, varying the sample lattice in the vertical and temporal dimensions, has been described.

The present invention has been described in terms of a preferred embodiment. The invention, however, is not limited to the embodiment depicted and described. For example, a frame rate of about 25 Hz with a corresponding field rate of about 50 Hz may be used instead of the frame rate of 30 Hz and field rate of 60 Hz used above. Rather, the scope of the invention is defined by the appended claims.

Claims

Claims: 1. A method of sampling an image having horizontal, vertical and temporal extent, comprising: sampling the image according to a sampling lattice running at a frame rate appropriate for magnetic tape recording, each frame of said sampling lattice including a first and second interleaved fields having uniformly spaced horizontal lines, the lines of one of said first and second fields coinciding with a subset of a set of horizontal lines of a corresponding field appropriate for said magnetic tape recording.

2. A method of sampling an image having horizontal, vertical and temporal extent, comprising the steps of providing a sampling lattice running at a frame rate appropriate for magnetic tape recording, each frame of said sampling lattice including a first and second interleaved fields having uniformly spaced horizontal lines, the lines of one of said first and second fields coinciding with a subset of a set of horizontal lines of a corresponding field appropriate for said magnetic tape recording; and sampling said image according to said lattice.

3. The method of Claim 1 or 2 wherein said subset comprises about one half of the lines of said set.

4. The method of Claim 1 or 2 wherein said frame rate is about 30 frames per second, and said first and second fields each comprise about 122 horizontal lines.

5. The method of Claim 1 or 2 wherein said horizontal lines are continuous.

6. The method of Claim 1 or 2 wherein said horizontal lines are linear arrays of sampling points.

7. The method of Claim 6 wherein each horizontal line comprises about 240 sampling points.

8. The method of Claim 6 wherein each sampling point contributes to a luminance signal and about one half of the sampling points contributes to each of the chrominance signals.