Physical and Visible Raster Areas

In the earlier discussion of how analog TV works, we saw that the picture was carried on a physical raster signal that had synchronization pulses and a defined area within which the picture is placed. Not all of the physical area is used for visible picture information, however. A little margin is provided at each end of the line scan, and a few lines are kept at the top and bottom as well, as those lines happen while the scanning circuits move the vertical position back to the top of the screen. That vertical repositioning is called flyback, and the stream of electrons to the CRT tube face is turned off while it happens. The time it takes for the flyback to happen is called the vertical blanking time.

So, for example, we have a physical raster of 480 lines with 720 samples. Converting an analog signal to a digital version sometimes leaves a small portion of black on either side of the picture. There are often some black lines at the top and bottom as well.

It is not necessary to compress these, but they must be there at the analog output if we are feeding a TV set or video recorder with our decompressed video. If we crop at the input stage, the playback system must ensure that the output physical raster is correctly sized.

If that is not done correctly, we will find our picture cropped more than we expect at the output stage. This is not fatal, but if we know about it, a safe margin placed around the image will prevent problems later.

Figure 6-6 illustrates how the physical raster is derived from the scanned TV signal.

6.9 D-Cinema

Digital cinema (D-cinema) is becoming more sophisticated and operationally easier to deploy. The IBC and NABconferences in 2003 and 2004 featured more coverage of this topic than ever before, with special screenings of full-length feature movies. This included a digitally processed version of the old Robin Hood movie starring Errol Flynn. The pro-cessing worked from the original Technicolor separations and combined them digitally.

Only a small amount of restorative work was done, but the vast improvement in quality came by working from a previous-generation master rather than simply digitizing a print that had been optically combined. With the advent of DVDs there has been a surge in demand for old movies to be restored and compressed using good-quality encoders in order to make them available to a new generation of consumers.

6.9.1 Dynamic Range

The imaging quality of the D-cinema format is improving all the time. When shooting this material with a completely digital video camera, the dynamic range of the content at the dark end of the intensity scale is still not as good as analog film. The detail is simply not there although it is improving year by year. Some compensation for this takes place in a film-to-D-cinema transfer. Where there is detail to be extracted, changing the gamma curve and adjusting the input-to-output intensity mapping will reveal subtle shading in dark-shadowed parts of the picture. Shooting must take place with strong lighting being carefully bounced into all the shaded areas of the scene.

Shortly, we will examine some of the work on high dynamic range imagery (HDRI).

This technique will lead to some significant improvements in the quality of digital pho-tography, especially where it pertains to low light conditions.

6.9.2 Film Grain

Cinematographers have fallen on either side of the fence regarding whether they like the quality of output produced with this technology. George Lucas clearly favors shooting

Action safe area Text safe area

Horizontal blanking period Overscan margin

Vertical blanking period

Physical raster lines

Figure 6-6 Safe areas.

straight to disk with a digital camera, while Steven Spielberg has gone on record to say he prefers the grainy texture of traditional film. Movies such as Minority Report achieve a good compromise by using digital effects to realize what the director envisioned while still preserving the film grain in the finished product by reintroducing it as a digital effect. As ever, there is clearly no right or wrong way to do this and the directors are likely to con-tinue making movies with a mixture of technologies for some time to come. From the point of view of a compressionist, film grain is a bad thing.

6.9.3 Image Sizes

With the work that movie special-effects companies have been doing on fully rendered 3D animation, a third alternative of completely computer-originated material is now avail-able. Two companies in particular represent the technological leading edge. Industrial Light & Magic (ILM) began work in this area in the 1970s. Some of the computer graphics work they were doing during the early 1980s was spun out to become the Pixar company.

Both have continued to push the state of the art forward ever since.

Several companies are looking at resolutions for D-cinema that go well beyond the current 1920 ×1080-sized pictures we are becoming accustomed to. Barco, Sony, NHK, and Olympus all have interesting research and development projects underway.

6.9.4 Aliasing Artifacts

Pixar has consistently delivered computer-animated content that has a vibrant and realistic image quality. Pixar representatives have explained at conferences how they use many tricks and shortcuts in order to simulate some computationally intensive techniques to deliver something that looks as if it required far more computing power than was actually used.

Movies for cinema presentation are typically created at a 2.35:1 wide-screen aspect ratio. The resolution of an individual frame of 3D animation is rendered at 2048 ×862 pix-els. The quality that Pixar produces appears to be that of a much higher resolution. This is because for every pixel being rendered, the computation averages an image that is drawn at a much higher resolution. This saves a lot of computation time when dealing with anti-aliasing artifacts. Figure 6-7 shows the effects of anti-aliasing (jaggies) on lines, and how they are alleviated by anti-aliasing, which is covered in the discussion on visible artifacts.

Table 6-5 D-Cinema Display Resolutions

Name Width Height

Panasonic large-venue DLP projector 1366 1080

HDTV/D-Cinema 1920 1080

Barco D-Cine DP100 2048 1080

Sony high-performance D-Cinema projector 4096 2160

NHK Ultra-HDTV prototype 8000 4000

The industry is settling on some standard screen sizes for D-cinema that coincide with the HDTV specifications, so imaging areas of 1920 ×1080 will become popular.

6.9.5 Resolvable Detail

You would think that a cinema screen would require more than 1080 lines to provide an image of sufficient quality. Reducing the pixel resolution to 1080 lines is feasible, though, because the audience only resolves about 900 lines at the normal viewing distance—even in a good theater. IMAX theaters are a special case because the viewer sits a lot closer to the screen than normal. Using the best equipment, such theaters achieve approximately 2000 lines of resolution on a 35mm print. Since the detail is just not visible, it is pointless to put the extra information in the output image. While the justification is there for 35mm to be printed at that quality, the potential savings in storage capacity and bit rate for D-cinema are more compelling.

Figure 6-8 shows how the resolving capabilities of the human eye limit the useful-ness of higher-resolution displays.

Aliased

Antialiased

Vector

Figure 6-7 Anti-aliasing effects.

Line 100

Line 101 1080 line image projection

Figure 6-8 D-cinema resolving power.

Some compression is taking place as the movie is laser scanned onto film. This can be disappointing to hear if you have labored over an 8K square image. Nevertheless, an image that is created and worked on at 8K is going to look a lot better in the final low-resolution output than if you had worked at the lower resolution, so the effort is not wasted.

Knowing about this compromise in resolving power significantly alters the cost dynamics of the disk storage you buy when equipping the distribution and play-out systems in a cinema complex. It also helps facilitate the delivery and transmission of a movie via a network connection from the studio to the theater. Video compression helps some more, too, but it does not have to compress the image quite as harshly to accomplish useful bit rates.

Delivering an appropriately sized image directly to a D-cinema projector eliminates the intermediate process of creating a 35mm print. This improves the quality of the view-ing experience for the customer and saves money for the distributor. Typical resolutions for this are 2200 × 1125 pixels, although 1920 ×1080 is also popular and is a size which coincidentally is also used for high-definition TV in some deployments.

6.9.6 Down-Sampling to SDTV Format

Let’s say you are creating some program material in HDTV format, for example, rendered at 1920 ×1080 and interlaced at 25 fps. If this is intended for delivery at standard definition for broadcasting purposes, you must render a special low-resolution version unless you trust the broadcasters to convert it themselves. The visible area on a 625-line raster is only 579 lines, which does not have an integer relationship to 1080. Even more problematic is the 480 visible lines on the 525-line systems, which have added complexity of scanning at 30 fps. The change in time base causes more problems with keeping the motion smooth.

This is especially tricky if the motion was rendered at field rate rather than frame rate. The interpolation artifacts are avoided by rendering two additional copies of the movie with the correct scale factor and frame timing. Most production companies will not have the luxury of time or computing capacity to do this so some shortcuts will be necessary.

6.9.7 Up-Sampling to HDTV Format

Time pressure being what it is, redesigning your rendering process could cut some corners off the time required to complete your project. For example, a raster only 720 lines high is probably detailed enough to be displayed at high definition. Up-sampling this 720-line image to 1080 lines is feasible for an HDTV output. A picture size that is 720 lines high is one of the HDTV standards currently deployed anyway. If you create content for multiple territories, then rendering at 25 fps will be less work than 30 fps. Pulldown will yield the 30 fps version as it does for movie film. Products such as Adobe After Effects will produce your up-sampled and down-sampled versions. The savings allow you to render 1080 ×720

×25p instead of 1920 ×1080 ×60i, which reduces your render workload to about 30% of what it would be at full resolution.

Compression is not just about bit rates coming out of a tool like Discreet Cleaner; it is a mindset. You have to think about your entire workflow process and not waste time creating data that is going to disappear when the video is compressed.