Digital Cinema's Special K

Is There an Economical Way to Bring 4K Processing Into the 2KPost World?

Figure 1: Standard scanning resolution for Academy, Scope, and Super35mm aperture in 2K and 4K.

In the standardization discussions for digital cinema distribution,there is an ongoing debate over how much resolution is required forpreparation, delivery, and display of theatrical images. How muchresolution is necessary to deliver higher quality than filmdistribution currently offers? What is possible, what is practical,what is necessary, and what is affordable? Is 4K resolution required atall stages, from capture to display, to preserve adequate imagequality?

The ultimate solution must satisfy a number of concerns. It mustmaintain the creative intent, provide better image quality thancurrently available, and it must present an economically compellingbusiness plan.

To understand the debate, let's first examine the background of 2K,4K, and resolution in general. For modern feature film production, withfew exceptions, digital images are processed at 2K resolution (seeFigure 1, next page).

A 2K workflow would normally mean that the original negative isscanned at 1828 pixels (for Academy or flat), is processed at thatresolution for effects, and written to film at 1828 pixels. Similarly,a 4K workflow would operate at 3656 pixels. In each case, output may befor digital projection, which means it would be resized to suit theresolution of the digital projector — for example, 2048pixels.

This is a contentious issue. Due to the transfer process, afilm-based workflow loses information at each step in the chain. Thisinformation loss starts in-camera, where lens and film movement reduceresolution of the images captured. The film itself has resolutionlimitations, and each stage of replication further losesinformation.

Figure 2: Note the aliasing on the fishing line and around the lampthat is present on the 2K scan, but not on the 4K. This indicates that4K scans are preserving information from the original that is notavailable on the 2K scans. (Images courtesy of Cintel InternationalLtd.)

Laboratory tests using carefully written test patterns on the filmdon't tell the whole story. From a practical approach, we can test thelens, camera, and film system by scanning an image and looking foraliasing, or jagged diagonal lines. If there are no such artifacts,then the sampling structure is more than twice the information contenton the film. If there are “jaggies,” then the informationcontent is close to (or greater than) the sample structure.

Figure 2 shows highly magnified 4K and 2K scans from the center of afull-frame 35mm image. Note the diagonal lines — very fineinformation on the lamp, fishing pole, and line. The 2K version of theshot shows alias artifacts, while the 4K scan shows smooth lines. Thisis an indication that the 2K sample structure is inadequate forsampling the image content, while the 4K is adequate. It is alsopossible to see these differences in the visual clarity and detailpresent in 4K and 2K scans.

Figure 3 (above), for instance, shows another image, shot on Kodak5245, low-speed, micro-fine grain, daylight negative film. Compare the2K and 4K scans, and you can see the additional details in theeyelashes and the increased sharpness in the eye visible in the 4Kversion. The structure of the grain on the film starts to becomeapparent in the 4K scan, though it is almost invisible in 2K. A 4K scandefinitely carries more information than a 2K scan.

We have seen that the original camera negative supports enough imageinformation to (technically) justify a 4K scan. But what happens asthis image travels through the processing chain?

Each time the image information is transferred to a different mediumit suffers from generation loss. This occurs because film is notcapable of fully capturing all the information from the scene at thehighest resolution. In fact, through the generations from negative toI/P to I/N to print, the contrast of the finer details diminishes untilthose details finally disappear. Technically, this is described asmodulation transfer function (MTF). Figure 4 (on page 38) shows atypical MTF curve for camera negative film.

Figure 3: Close-up of 2K and 4K scans. Note that the 4K scan carriesmore information and detail than the 2K scan, especially in theeyelashes and the overall sharpness of the eye. Also note the increasedgrain visible in the 4K scan. (Images courtesy of Cintel InternationalLtd.)

It is important to understand that MTF is multiplicative throughprocess steps. When two processes are cascaded, the resulting MTF isthe convolution (or multiplication) of the curves. Fundamentally, thismeans that each process will further degrade the modulation of finedetail, causing it to disappear sooner.

As an example of this, look at Figure 5. This is a comparison of 4Kscans of original negative and an interpositive. The I/P has beencontact-printed from the original negative. Note, in particular, thatthe scan of the I/P is noticeably softer. This is the result ofgeneration loss. Additional loss is incurred at each step as the filmis printed from I/P to I/N to release print.

A technical analysis can model the entire process from the scene tothe screen and determine the actual amount of information that willreach the screen. The net effect of this cascading effect is that thereis value in having one or more higher resolution steps in theprocessing chain.

Of particular interest in the MTF curve are the mid levels of detailbetween 10 and 20 lp/mm on film. This band carries the information thatthe human visual system is most sensitive to, and therefore, mostinterested in. If the contrast in this region is higher (as shown bythe curve being higher), then the image will appear sharper, withoutactually having higher resolution (see Figure 4).

Generation loss and MTF degradation in the film postproduction chainresult in projected film images with substantially lower performancethan those same images captured on the original negative. There havebeen a number of studies done that support this conclusion,illustrating that the information contained on typical release printsis only some of what was contained on the original negative.

We have seen that there is definitely more information carried in a4K resolution image than a 2K image although, for many images, thedifference is subtle.

Figure 4: Typical MTF curve for Kodak 5218 film (brown curve), andthe system MTF curve resulting from processing an image through acamera lens, then onward to intermediate stock, and eventually torelease print. Note that the contrast of the mid-resolutions (20 lp/mmto 60 lp/mm) has significantly reduced from the system point of viewfrom the original negative. (Chart created by Entertainment TechnologyConsultants.)

Thus, while working at higher resolutions clearly offers bettervisual performance, it does so with diminishing economic returns forthe extended effort. Since a 4K image contains four times as manypixels as a 2K image, with a proportional increase in scanning,rendering, and film output times, as well as four times as much storagerequired for the raw and final data, a significant cost multiplierexists when embarking on a 4K post process. This increased cost resultsin higher postproduction budgets, and longer schedule time for managingand rendering 4K data, instead of 2K.

The economics of 4K postproduction processing are likely to improveas processing power increases and disk storage costs drop in thefuture. Still, in the current film environment, almost all digital filmwork is being done at 2K, mainly due to the time and economicconstraints associated with 4K work. In other words, the improvedperformance is not generally considered to be worth the cost by theindustry's financial gurus.

How does this affect the digital cinema debate? Obviously, intoday's production environment, the cost/performance trade-off hassettled on 2K as the standard resolution for production of featurefilms. The current argument suggests that digital projectors arecapable of maintaining 2K resolution right to the screen, andtherefore, this is a superior approach to current release printperformance.

Figure 5: Scans of original negative and scans of corresponding I/P.Note the generation loss apparent in the print from negative to I/P.(Images courtesy of Cintel International Ltd.)

Thus, the debate is about enhancing image quality for the future,not about maintaining the status quo.

The economic hurdles largely revolve around the cost of 4Kprojectors and developing 4K theater infrastructures. Certainly,developing a 4K projector is difficult. R&D investments to scale upfrom 2K to 4K are currently huge. And right now, these investments aresupported by a fairly small potential market of worldwide moviescreens. This class of performance is not very interesting to themultimillion unit consumer markets, and so they hope to amortize theR&D over much smaller volumes. This, in turn, renders the 4Kprojector economically unattractive.

Similarly for the 4K in-theater infrastructure of servers and highbandwidth data links between servers and projectors, the hardwarerequirements to support 4K are large, and the cost of supportingrealtime 4K image bandwidth to the projector is prohibitive.

It is useful to look to the film world for guidance on the issue ofeconomic trade-offs. It is well known, for instance,that 70mm producesa superior on-screen image to 35mm, but the economics and practicalaspects of producing and projecting in 70mm have kept the format out ofthe mainstream.

Thus, the question is whether 2K is good enough, and if so, how tomaximize its performance through appropriate postproduction techniques.Is there a compromise that will satisfy both the drive for higherquality, and the economic realities?

This brings us back to the 2K vs. 4K argument. We have seen thatthere are visual merits in maintaining high resolution in one or morestages of the image-processing chain, even if the others arelimiting.

Figure 6: Comparison of 2K and 4K origination as displayed on a 2Kprojector (simulated). The 2K scan represents a 2K image directlyprojected from 2K data. The image on the far right represents a 4Kimage downconverted and projected at 2K. (Images courtesy of CintelInternational Ltd.)

It is instructive to compare the visual quality that would resultfrom a 2K scan-process-display pipeline to a 4K pipeline. Generally, wewould see some visual benefits from the 4K pipeline, as demonstrated inthe images that accompany this article. Those benefits would be small,but apparent to the well-trained viewer in an optimal environment.

Economic issues therefore enter the argument. It is extremelyunlikely that a 4K projector and server system will be economicallyavailable to the cinema market in the near term. 2K performancerepresents the best available for some time to come.

Does this mean that digital cinema cannot move forward? Notexactly.

Recent demonstrations of 2K resolution imagery from TexasInstruments and Kodak have shown resolution and image detail that isbetter than what 35mm film prints deliver to cinemas today. These 2Kprojectors, however, can be made to look even better.

These demos were done using postproduction workflows that scannedand processed at 1920 resolution with sub-sampled color. Is there a wayto improve on this?

Because the image degradation is multiplicative step-by-step, if oneor more steps can be performed at higher quality, visual benefits willresult. Consider Figure 6: Both images are shown at 2K resolution, butone was scanned and processed at 4K, the other at 2K. Note thedifference — while subtle, the 4K-scan-originated image clearlypreserves more of the original information.

The solution to the question of how to develop cost-effective 2K/4Kpostprocessing systems involves complex trade-offs between absoluteimage quality, practical economics, and the determination of anindustry-standardized acceptable level of performance.

Clearly, a 4K end-to-end system would be preferable to a 2K system,much in the same way that 70mm would be preferable to 35mm in film, ifit were practical.

Unfortunately, this would render digital cinema cost-prohibitive.The current 2K prototype projectors being tested around the countryshow that 2K digital pictures projected on a large screen areoutstanding and already capable of offering a better picture to payingaudiences.

The 2K vs. 4K debate thus requires a practical solution: Mixing 4Kand 2K processes will yield higher quality results than a 2K workflowalone. Scanning and postproduction processes done at 4K and convertedto 2K for distribution to theaters will provide outstanding results ona high-quality 2K projector.

A compromise solution of 4K capture and processing, distributed at2K resolution and playing back on a 2K projector, will yield superiorvisual images to a 2K end-to-end system, with the financial effectsoccurring in the mastering stage, not at every theatricalinstallation.

Matthew Cowan is a principal at Entertainment TechnologyConsultants, an organization specializing in the science andapplications of digital cinema technology. He has more than 20 yearsexperience in the development and application of new products in themedia and display fields.

His background includes development of electronic projectionsystems, analysis of color reproduction issues in electronic displays,strategic technology sourcing, reviews of advanced electronicprojection products, and detailed analysis of compression schemes fordigital images. Cowan was instrumental in developing the currentmastering processes used in digital cinema, which introduced the use ofthe digital mastering theater for color and dynamic rangeadjustment.

The author would like to thank Peter Swinson for discussions andinsight into 4K scanning, and Cintel International Ltd. (www.cintel.co.uk)for kindly providing an excellent set of images for thisarticle.