Video resolution

Too many lines would be a waste and too few lines would make the raster line structure visible. A decision had to be made. Camera sampling and picture
Author:
Publish date:

Too many lines would be a waste and too few lines would make the raster line structure visible. A decision had to be made.

Camera sampling and picture display suffer from vertical uncertainty irrespective of whether the scanning is interlaced or progressive.

The extent to which a picture medium such as television can reproduce fine detail is expressed in terms of resolution. The early development of television resulted in the two dominant SDTV scanning formats, the 525/60 and the 625/50. The aim was to achieve a satisfactory picture taking as a reference the visual acuity of the eye. The human visual system (HVS) has two main resolution characteristics namely: The spatial resolution and the temporal resolution.

The spatial resolution concept

Television system design takes as a reference the visual acuity of the eye, which is of the order of one minute of arc. The eye does not perceive picture details that subtend an angle of less than one minute of arc. The assumption was made that the picture would be viewed at a distance of approximately six times the screen height. So a decision had to be made as to the number of lines that make up a picture. Too many lines would be a waste and too few lines would make the raster line structure visible. North America chose 525 lines and Europe chose 625 lines.

FRAME GRAB
A look at tomorrow's technology

Summer hits Internet Top 10
Internet usage slows as schools close

netScore Top 50 Properties - June 2001 June 2001
Rank Property June 2001
U.S. Unique
Visitors (000) % Change
May-June 1 Microsoft Corporation 91,341 -5.4% 2 Yahoo Inc. 86,465 -5.5% 3 AOL Time Warner 80,256 -2.5% 4 Terra Lycos S.A. 49,949 -9.1% 5 At Home Corporation 35,880 -9.4% 6 Amazon.com Inc. 30,118 4.1% 7 U.S. Government 29,655 25.4% 8 InfoSpace Inc. 28,941 -12.8% 9 Primedia Inc. 28,593 -9.9% SOURCE: comScore Networks & Diameter
www.comscore.com, www.diameter.net

  • The vertical resolution: The vertical resolution is equal to the number of alternately white and black horizontal lines that can be resolved vertically over the full height of the picture. It is expressed in lines per picture height. The 525/60 and 625/50 scanning standards use interlaced scanning. The early developers of television in the 1930s determined that in an interlaced scanning system the vertical resolution is statistically equal to 70 percent of the number of active lines. The so-called “Kell Factor” of 0.7 is at the origin of all conventional television systems. Given an active number of lines, let's say 485, the vertical resolution is equal to 0.7×485=339.5 so let's round it up to 339 lines per picture height (LPH). If the vertical details in the picture exceed 339LPH they are blurred. The vertical resolution is independent of the video bandwidth. Format Total lines
    per picture Active lines
    per picture LPH (K=0.7) Bandwidth
    (MHz) Lines per
    MHz 525/60
    (CCIR M) 525 485 339 4.2 79.2 625/50
    (CCIR B, G) 625 575 402 5 80.4 Table 1. Horizontal resolution capability of the dominant SDTV formats.
  • The horizontal resolution: The aim is to achieve an equal number of picture elements per unit of distance (the picture height) horizontally as well as vertically. The system must therefore allow for a number of horizontally displayed picture elements that is equal to the picture ratio, let's say 4/3, multiplied by LPH. So in the 525/60 format we need to display 339×4/3=452 picture elements (or 226 cycles) per active picture width. Given an active line duration of 52.5 microseconds this results in a cycle duration of:
    T = 52.85/226 = 0.2338 microseconds

The associated frequency is
F = 1/T = 1/0.2338 » 4.28MHz

This is the bandwidth required for equal horizontal and vertical resolution. The horizontal resolution factor for a 4.28MHz bandwidth is 339/4.28 = 79.2 lines/MHz

Given that the transmitted bandwidth is 4.2MHz the transmitted horizontal resolution is reduced to:

4.2MHz × 79.2 lines/MHz » 333 lines

Reducing this bandwidth reduces the horizontal resolution. A 2MHz bandwidth, typical of VHS, would result in a horizontal resolution of about 160LPH without affecting the vertical resolution. Table 1 compares the resolution capabilities of the 525/60 and 625/50 scanning formats.

Format 525/59.94 625/50 Component E'Y E'CB E'CR E'Y E'CB E'CR Sampling frequency (MHz) 13.5 6.75 6.75 13.5 6.75 6.75 Nyquist frequency (MHz) 6.75 3.375 3.375 6.75 3.375 3.375 LPF cutoff frequency (MHz) 5.75 2.75 2.75 5.75 2.75 2.75 Horizontal resolution (LPH) ≈455 ≈218 ≈218 ≈449 ≈215 ≈215 Samples per total line 858 429 429 864 432 432 Samples per active line 720 360 360 720 360 360 Table 2. 4:2:2 sampling structures and horizontal resolution.

NTSC and PAL composite signal encoding formats have a typical color difference (B-Y and R-Y) bandwidth of the order with a resulting horizontal resolution of around 48LPH.

The temporal resolution concept

An important property of the eye is persistence of vision. Persistence of vision is the ability of the viewer to retain, or in some manner to remember, the impression of an image after it has been withdrawn from view. When light entering the eye is shut off, the impression of light persists for about 0.1 sec. Ten still pictures per second is an adequate rate to convey the illusion of motion. The phenomenon of flicker, however, requires still higher picture rates. Given the transmission spectrum conservation requirements, which imposed a 6MHz channel bandwidth (7- or 8MHz in Europe), the early television developers adopted the interlaced scanning concept, in which a picture is divided in two consecutive fields transmitted at a frequency (nominally) 60Hz (50Hz in Europe). This allowed for the reduction of the transmission bandwidth requirements. The result is that large areas of uniform color and brightness flicker at the field rate (60Hz or 50Hz large area flicker). This is an acceptable compromise. When two adjacent fields in two consecutive fields have different luminance values the result is small area flicker, at the frame rate (30Hz or 25Hz), which is highly objectionable. In the 1930s this was a small price to pay to achieve restricted transmission bandwidth.

The ITU-R BT.601 digital concepts

The ITU-R BT.601 standard was the first international agreement on how to migrate from two incompatible analog television scanning formats to a common sampling concept and structure. The original analog component signals are sampled to obtain three digital component signals. The most popular sampling structure, the 4:2:2, uses a 13.5MHz sampling rate for the E'Y (analog luminance) signal and 6.75MHz for each of the E'CB (blue color difference) and E'CR (red color difference) component analog signals. The sampling frequencies impose Nyquist constraints on the maximum sampled analog frequency that has to be lower than half the sampling frequency to avoid the occurrence of aliasing. The standard specifies the tolerances of the anti-aliasing and reconstruction filters thus implicitly specifying the analog horizontal resolution.

In the 525/60 version the luminance sampling frequency of 13.5MHz = 858 × FH, where FH is the horizontal scanning frequency. The resulting number of samples per total line is equal to 858. The digital active line accommodates 720 Y samples (active pixels). Under ideal conditions, given the Nyquist frequency of 6.75MHz, 720 pixels per active line is equivalent to 3/4 × 720 = 540LPH. The standard specifies an anti-aliasing and reconstruction filter bandpass of 5.75MHz resulting in an analog horizontal resolution of the order of 455LPH.

Given the color-difference sampling frequency of 6.75MHz, the digital active line accommodates 360 CB and 360 CR samples (active pixels). Under ideal conditions, given the Nyquist frequency of 3.375MHz, 360 pixels per active line is equivalent to 3/4 × 360 = 270LPH. The standard specifies an anti-aliasing and reconstruction filter bandpass of 2.75MHz resulting in an analog horizontal resolution of the order of 218LPH.

The analog resolution figures above assume ideal brickwall lowpass filters. Such filters don't exist in practice. Under the influence of the computer industry, various bodies have started referring to the number of samples (pixels) per active line as horizontal resolution and to the number of active lines per picture as vertical resolution. This is misleading. Table 2 summarizes the situation.

ATSC concepts

The ATSC standard picks up where “601” left off and applies the principles of 4:2:2 sampling to SDTV as well as HDTV scanning formats. The confusion starts with the 4:2:2 nomenclature, which originally meant 4×3.375MHz=13.5MHz and 2×3.375MHz=6.75MHz. The normalized HDTV sampling frequencies of 74.25MHz for Y and 37.125MHz for CB and CR would suggest that the sampling structure be called 22:11:11. This was however never considered. The confusion is carried further by the coexistence of two HDTV formats: an interlaced scan format of 1125 lines per picture (1080 active lines) and a progressive scan format of 750 lines (720 active lines). Comparing the resolution capabilities of these two formats and expressing them in an easily interpreted number is quite a challenge.

The specified antialiasing and reconstruction filters create the same horizontal resolution ambiguities. Here the analog related horizontal and vertical resolution concepts are completely ignored and the quoted resolutions are the active horizontal pixels and the number of active scanning lines. It is interesting to note that the vertical resolution uncertainty that has produced the Kell factor is currently criticized or completely ignored. Whether the generally used figure of 0.7 is realistic or not and whether it applies equally to interlaced and progressive scanning is immaterial. The fact is that camera sampling and picture monitor display suffer from a certain amount of vertical uncertainty that occurs irrespective of whether the scanning is interlaced or progressive. While the uncertainty is reduced with progressive scanning the fact remains that unless the signal source is a computer the vertical resolution is not equal to the number of active lines. Also using the number of samples per active line to express horizontal resolution is misleading.

Michael Robin, former engineer with the Canadian Broadcasting Corporation engineering headquarters, is an independent broadcast consultant in Montreal, Canada. He is co-author of Digital Television Fundamentals, published by McGraw-Hill.

Send questions and comments to: michael_robin@intertec.com

The Second Edition of Michael Robin's book may be ordered directly from the publisher by calling 800-262-4729. It is also available from several booksellers.