The ATSC/DTV standard specifies two picture aspect ratios (4:3 and 16:9). The 4:3 aspect ratio formats are, respectively, 720×480 and 640×480. The 16:9 aspect ratio formats are 1920×1080, 1280×720 and 720×480.
A period of transition, originally slated to last until the end of 2006, is expected to allow the gradual implementation of 16:9 aspect ratio DTV transmissions. During this period TV stations are expected to simulcast 4:3 aspect ratio NTSC analog signals on the currently allocated channels and 16:9 aspect ratio DTV transmissions on separate newly allocated terrestrial transmission channels. The simulcasting is planned to stop in the year 2006, or later, when all analog NTSC transmissions are expected to end and the related transmission channels assigned to other uses. In the transition period a great deal of format conversions, mostly 4:3 to 16:9, will take place.
There are three methods of 4:3 to 16:9 format conversion:
The top and bottom crop mode: Figure 1 shows the manner in which a 4:3 aspect ratio picture is stretched in the horizontal and vertical direction to fill a 16:9 aspect ratio screen, resulting in a 25 percent loss of vertical resolution. The viewing window can be preset or a tilt-and-scan approach can be used. Here the operator moves the window in the vertical direction to follow the action.
The side panel mode: Figure 2 (on p. 34) shows the manner in which the original 4:3 aspect ratio picture is inserted in a 16:9 window, resulting in black side panels. The side panels occupy 25 percent of the horizontal space.
The anamorphic distortion mode: Figure 3 (on p. 34) shows the manner in which a 4:3 aspect ratio picture is stretched horizontally to fill a 16:9 aspect ratio screen, resulting in a 33 percent anamorphic distortion.
None of the three methods offers an ideal solution. Experiments indicate that a 5 percent anamorphic distortion is undetectable and a 7 percent anamorphic distortion is not objectionable. The current trend is to combine the three methods to obtain a picture that is subjectively pleasant to the viewer.
Resizing changes the original size of an image, i.e. the number of pixels used to represent the image, to an arbitrary target size. The process effectively is a change of the sampling rate of the image. Therefore resizing and resampling digital video are the same process. The image resizing has two aspects as follows:
Upscaling: Generating a higher number of scanning lines (NV) and/or horizontal pixels (NH) by a process of interpolation. The interpolation simulates image positions that do not correspond to available samples.
Downscaling: Reducing the number of scanning lines (NV) and/or horizontal pixels (NH) by a process of decimation.
Like all processing of images in the digital domain the interpolation is achieved by performing mathematical operations on the number of samples representing the original image. In typical applications the mathematical operation is a multiplication complemented by a division. In the analog domain this multiplication process is performed by an analog multiplier and is a continuous process. In the digital domain a digital multiplier operating on each sample of the image performs it. Resizing the image affects the sampled spectrum, resulting in aliasing. Finite-impulse-response (FIR) digital filters are used to avoid aliasing.
Figure 4 describes the horizontal and vertical upscaling process. The assumed aspect ratio conversion uses the vertical cropping concept. The conceptual block diagram consists of three basic elements: the upsampler, the FIR filter and the downsampler.
As an example of upscaling assume an ITU-R601 original picture format with the following characteristics:
NV: 360 after vertical cropping (480 × 0.75 = 360)
Assume a desired picture format with the following characteristics:
The ratio of N'H/NH = 1920/720 = 2.666… = 8/3
The horizontal interpolation consists in upsampling NH by a factor of 8, effecting a FIR filtering and downsampling the result by a factor of 3.
A similar process is effected in the vertical direction resulting in a ratio of:
N'V/NV = 1080/360 = 3
In this case the interpolation consists of upsampling NV by a factor of three, effecting a FIR filtering and downsampling the result by a factor of one.
Other conversion ratios will be used for upscaling to the 1280×720 format. Table 1 details the upsampling (↑) and downsampling (↓) multiplying factors used to convert a 720×480 picture format to 1280×720 and 1920×1080 picture formats using either the vertical cropping (tilt-and-scan) or the side panel method. Note that in the vertical cropping mode the number of active lines in the original picture is reduced to 360. In the side panel mode the number of active pixels in the HDTV 16:9 formats is reduced from 1280 to 960 and from 1920 to 1440 respectively.
Upconversions will be used in the short run to modify 4:3 aspect ratio SDTV-originated local programs for insertion into 16:9 aspect ratio DTV network feeds. In the long run, upconversion will be used to convert 40 years of SDTV legacy programming to DTV formats. Figure 5 shows the conceptual block diagram of an upconverter. It has several basic functions:
Serial-to-parallel conversion (S/P): It is assumed here that the input signal is ITU-R601 bit-serial with a bit rate of 270Mb/s. If this is not the case then the source signal format has to be suitably converted. In any case it is preferable to use digital component signal sources to avoid NTSC artifacts.
Image processing: The image processing function is essentially a picture cleanup process aiming at reducing the noise and removing NTSC (e.g. crosscolor) and MPEG artifacts introduced by previous signal encoding, decoding and distribution processes.
Frame synchronizer (FS)
The deinterlacer (DI): The interlaced-to-progressive conversion transforms interlaced fields into progressively scanned frames. Calculating the “missing” lines in an interlaced field accomplishes this. Sophisticated techniques require field memories to avoid “judder” effects resulting from motion in the image.
The aspect ratio converter (ARC): This function implements the type of picture format modification such as vertical cropping with tilt-and-scan, side panel or anamorphic distortion. Each of these approaches has its disadvantages.
Horizontal upscaling: Generating a higher number of horizontal pixels (NH)
Vertical upscaling: Generating a higher number of scanning lines (NV)
Parallel-to-serial converter (P/S): Conversion to SMPTE 292M
The term downconversion applies to hardware that reduces full HDTV image data to form an SDTV resolution picture. The process of reducing the sampling rate is called decimation. In our May 2001 article we discussed the three 16:9 to 4:3 aspect ratio conversion methods, namely the horizontal cropping, the letterbox and the anamorphic distortion. In the digital domain these aspect ratio conversion methods will use downsampling.
Downconversion will accommodate viewers using 4:3 aspect ratio television receivers and watching 16:9 aspect ratio originated pictures. NTSC stations will use downconverters to feed the NTSC transmitter with 16:9 originated signals. In addition, given the large number of analog 4:3 format NTSC receivers in use, it is unrealistic to expect that all these receivers will be discarded in 2006. In all likelihood a large number of set-top converters/decoders will be used to convert the 16:9 DTV transmissions to feed the 4:3 NTSC analog receivers. These set-top decoders will use some type of format downconversion.
Figure 6 shows a simplified conceptual block diagram of a downconverter.
Michael Robin, former engineer with the Canadian Broadcasting Corporation's engineering headquarters, is an independent broadcast consultant located in Montreal, Canada. He is co-author of Digital Television Fundamentals, published by McGraw Hill.
A look at tomorrow's technology
More than one in four consumers prefer TV to PC for shopping.
|18-34 year old||25%|
|50+ year old||42%|
SOURCE: Pace Micro Technology www.pacemicro.com
Send questions and comments to: firstname.lastname@example.org