Transition to Digital

Resolution Many of the aspects of today's legacy television standards (SDTV) were developed through experiments carried out in the 1930s. They reflect
Author:
Publish date:
Social count:
0

Resolution Many of the aspects of today's legacy television standards (SDTV) were developed through experiments carried out in the 1930s. They reflect the understanding of the psycho-physical perception of images applied to picture capture, efficient and economical signal transmission, and acceptable picture reproduction in the home, given the technology of the times. The extent to which a picture reproduction medium can reproduce fine detail and/or movement is expressed in terms of resolution. Historically resolution of a fixed image was understood to mean "limiting resolution," or the point at which adjacent picture elements of an image cease to be distinguished. Various disciplines measure and specify resolution differently.

Spatial resolution A photograph is essentially a two-dimensional representation of an event frozen in time. The difference between a photograph and motion pictures lies in the fact that "motion pictures" are a sequence of successive frozen images. The capability of the photograph to reproduce fine detail is expressed as the number of separate lines per millimeter (spatial resolution) which are just recognizable. Values can range from below 50 to above 1000. Photographic resolution depends on the emulsion, the contrast of the subject, the color of the light and other factors. The resolution of a photograph is the same in all directions.

A television picture is an optical illusion. The original picture is scanned horizontally and vertically fast enough to create an impression of continuity. The result of the picture scanning process is an electrical representation (in volts) of the picture brightness values (in candelas per square meter) at specific locations on the scanned image at a given moment in time. A television picture is made up of a given number of lines, specific to the scanning format. The SDTV standards (525/60 and 625/50) were developed taking into consideration the acuity of vision of the human eye (1 minute of arc), assumed viewing conditions (viewing distance six times the picture height) and transmission-spectrum savings concerns. Experiments indicated that the optimal number of scanning lines making up a television picture given these constraints is about 572. This ballpark figure is at the origin of the number of lines specified for the two SDTV standards - namely the 525-line system used in North America and Japan and the 625-line system used elsewhere in the world. A high-definition standard with 1125 lines per picture requires shorter viewing distances, e.g. three times the picture height, to enable the eye to resolve all picture details. Due to the horizontal and vertical scanning process, television pictures exhibit two types of spatial resolution, namely vertical resolution and horizontal resolution.

Vertical resolution is related to, but not equal to, the number of active scanning lines. In interlaced television, the vertical resolution is equal to the number of active lines multiplied by the Kell factor, usually taken to be 0.7. The resulting "statistical" vertical resolution is expressed in lines per picture height (LPH) and is independent of the transmission bandwidth. In the 525/60 scanning standard, the vertical resolution is equal to 0.7 x 485 = 339LPH.

For equal horizontal and vertical resolution, the number of picture elements (pixels) per active line duration that the television picture needs to resolve horizontally is equal to the aspect ratio (e.g. 4:3 or 16:9) multiplied by the vertical resolution. In the SDTV NTSC system this is equal to 339 x (4/3) = 452. Scanning 452 pixels results in an electrical signal with 226 complete cycles during the active horizontal scanning line. Figure 1 shows how scanning horizontally alternate white and black pixels results in a sinewave. Given an active line duration of 52.85sec, one complete cycle has a duration of 0.2338sec resulting in a maximum video frequency of Fmax = 1/(0.2338sec) = 4.28MHz.

The FCC permitted maximum transmitted frequency is 4.2MHz resulting in a slight loss of horizontal resolution. This bandwidth needs to handled by the transmission medium, all the way from the camera output to the TV receiver's CRT. The horizontal resolution is expressed in LPH and in an optimized system is equal to the vertical resolution. Every television format has a related horizontal resolution factor expressed as lines/MHz. In the SDTV formats (NTSC and PAL) the horizontal resolution factor is about 80 lines/MHz. A 4.2MHz bandwidth therefore results in an analog horizontal resolution of 80 x 4.2 = 336LPH, almost equal to the vertical resolution.

Temporal resolution An important property of the eye is persistence of vision. Once an image has been formed on the retina, the eye acts as a storage device and the visual sensation of the image is retained for a finite length of time. As a consequence, the human vision system (HVS) can be "fooled" into believing that a sequence of static pictures represents motion. This occurs if the display rate is greater than 10 successive pictures per second. A different phenomenon, known as flicker, requires still higher picture rates. The number of pictures per second is chosen to provide a sufficiently rapid succession to avoid display flicker at levels of image brightness appropriate for given viewing environments. The "critical flicker frequency" (CFF) depends on the display brightness and is the minimum rate of display, at a given picture brightness, at which the HVS does not perceive flicker. Early movies used 17 pictures per second. Contemporary movies use 24 pictures per second and each picture is projected twice, resulting in a display rate of 48 pictures per second. The television picture (frame) has a refresh rate of (nominally) 30Hz (525/60 scanning standard) or 25Hz (625/50 scanning standard). Each frame is made up of two successive interlaced fields with a refresh rate of 60Hz or 50Hz. Historically these frequencies were related to the power line frequency but today they are only nominally related to it.

The use of interlace is the consequence of the need to reduce the transmitted bandwidth and can be viewed as an early compression process. It causes large picture areas of uniform color and brightness to flicker at the field rate (large area flicker) and is an acceptable compromise. When two adjacent lines in two consecutive fields have different luminance values, the result is small area flicker at the frame rate, and this is highly objectionable. To avoid these problems, computers, which have no wide-bandwidth signal distribution constraints, use progressive scanning, a high number of lines and picture refresh rates in excess of 60Hz. Table 1 gives the CFF for commonly encountered flicker frequencies.

Digital system considerations The advent of digital signal processing has introduced a new twist in the concept of resolution. According to Nyquist theory, to avoid the occurrence of aliasing, the maximum sampled analog video frequency has to be less than half of the sampling frequency. With one exception, the composite digital 4Fsc standard, all digital video formats use component video signals. The analog component video signals E'Y, E'CB and E'CR are sampled at a multiple of the horizontal scanning frequency (FH) resulting in the pervasive 4:2:2 sampling strategy. There is, therefore, a direct relationship between the number of samples (pixels) per line and the digital sampling frequency. SDTV and DTV standards specify the characteristics and tolerances of the anti-aliasing and reconstruction filters. Table 2 shows details of three digital sampling structures.

The ITU.R 601 standard is based on the existing traditional SDTV formats: NTSC and PAL. The sampling frequencies adopted were a compromise aimed at satisfying NTSC and PAL users. Although it allows many sampling structures the most popular one is the 4:2:2 format. The low-pass filter characteristics show a moderate attenuation at half the sampling frequency. Applying analog resolution concepts, a low-pass filter with a cut-off frequency of 6.75MHz would result in a horizontal resolution of 540 LPH; the equivalent of 720 pixels per active line. High-energy luminance frequency domain components around 6.75MHz, even if only 12dB attenuated as allowed in the standard, stress the A/D process and generate aliasing. Practical filters limit the Y bandpass to 5.75MHz resulting in a horizontal resolution of 460LPH. Similar considerations apply to the two color-difference signals. In both cases the analog horizontal resolution is superior to NTSC and PAL but it still bears the footprint of analog low-pass filtering.

The SMPTE 296M standard describes a progressive scan HDTV system with an active raster of 1280 active horizontal pixels by 720 active lines. The format is not interlaced, so the interlace-related vertical resolution ambiguity, partly responsible for the Kell factor, is considerably reduced here. There are no clear guidelines aimed at specifying the statistical vertical resolution of a progressive scan television system. The maximum video frequency resulting from scanning 1280 active horizontal Y pixels is on the order of 37MHz, thus tightly matching the constraints imposed by the 74.25MHz sampling frequency. Again analog considerations would require a higher sampling frequency for optimal analog resolution. Similar considerations apply to the color-difference signals. Generally all analog resolution concepts are ignored and this format is described as having a resolution of 1280 x 720.

The SMPTE 274M standard describes an interlaced scan HDTV system with an active raster of 1920 active horizontal pixels by 1080 active lines. This format is interlaced. Consequently, the Kell statistical factor applies and the analog vertical resolution is 0.7 x 1080 = 756LPH. The maximum Y video frequency is again on the order of 37MHz and a 74.25MHz sampling frequency is slightly marginal. Similar considerations apply to the color-difference signals. Once again all analog resolution concepts are ignored and this format is generally described as having a resolution of 1920 x 1080.

As shown, in all digital formats the sampling frequency imposes constraints affecting the anti-aliasing and reconstruction filters resulting in a reduction of the horizontal resolution to avoid aliasing. These constraints obviously do not exist with digitally generated pictures. However, remember that nature is analog so the transmission of real pictures requires filtering and A/D conversion. With digital television concepts and implementations, the trend is toward expressing the picture resolution as the number of pixels per active line multiplied by the number of active lines per field. This trend tends to ignore the original definitions of horizontal and vertical resolution gradually replacing them in equipment and system specifications.