Comparison of HDTV and Motion Picture Film - Limiting resolution

"How to get the best possible image data for motion pictures?” – That’s a big question as we enter a time when digital projection increasingly will join film projection on cinema screens. Can film look better? Can digital challenge film for image quality on the big screen? If the goal is to improve today’s cinema experience and provide the best possible image master for the movies of the future, scientists need to know how to get the best possible image data.

To create film prints as well as digital copies -- all with consistency of color and to the highest standards of quality -- that data must exist in digital form. But, should the original images be captured with digital or analog technology? To find an answer, Kodak research scientists compared the performance of HDTV cameras and the scanner output from Super 35 mm Motion Picture film exposed in Motion Picture Cameras.

These assessments of film and electronic capture used both digitally sharpened images as well as images that were not sharpened.

Kodak scientists used Kodak Vision 500T color negative 5279 film at full aperture 35 mm (sometimes referred to as Super 35). Once developed, the film negative was scanned at 4K pixels per line with a Cineon Genesis scanner to produce digital data using a full scan resolution of 4096 pixels by 3072 lines. All three RGB color channels were at full resolution. Kodak software was then used to reduce grain, sharpen the image, and transform the image data into a common printing space. Because the test was standardized to the 1.85 format preferred by cinematographers, the image was cropped, choosing the most creatively appropriate region, to 4096 pixels by 2214 scan lines while maintaining 10 bits per color in each channel through the entire process. The images were then analyzed and viewed, and the results verified, using multiple images and approaches.

HDTV images at 1.85 format used a resolution of 1920 pixels per scan line. Unsharpened images were obtained using a 24P electronic camera set to zero sharpening, and the resulting image data was analyzed without applying any additional sharpening or peaking. A 24P HDTV camera that had the most boosted frequency response of the three 24P cameras was selected to provide sharpened HDTV images.

A finite impulse response (FIR) filter was used to process the 500T film image data. Ideally, the film would have been sharpened to reach the same artifact level as in the 24P camera. However, the sharpening filter used produced film data that had lower artifacts than the HDTV camera, while providing a sharper image.

A variety of metrics were chosen for analysis including Limiting Resolution that measures how much detail or information can be captured from the scene.

Scenes and targets appropriate to the metrics selected for analysis were chosen. For limiting resolution, sharpness, and aliasing metrics, the ISO 12233 standard target was used.

An enlargement of the frequency sweep area of one of the test charts used shows the differences in response. This pattern can be thought of as a way to show what would happen if a pattern of nine dark areas on a white background were zoomed out or drawn away from the camera, while remaining in focus. A similar effect can also occur if an object with a gray regions turns away from the camera, while staying within focus. The four conditions are (a) film unsharpened, (b) film sharpened, (c) HDTV camcorder set to no detail or sharpening enhancement (d) HDTV camcorder with the most boosted response of the HDTV cameras tested.

The red arrows correspond to the limiting resolution point of each image. The ISO standard 12233 associated with this test chart defines limiting resolution as "the value, in line widths per pictures height (LW/PH), of that portion of the wedge where the resolution response (average depth of modulation value) equals 5% of the reference response."

The plot of frequency response shows the limiting resolution at the 5% level for each of the four conditions.

Note the significantly higher limiting resolutions for the film capture system compared to the HDTV capture system.

How does consumer film differ from motion picture film results? One difference is that Kodak uses different film types for consumer film than motion picture films. Another important difference is that the area of an 8 perf consumer film frame is 2.2 times the area of a Super 35 1.85 frame and so potentially because of the larger area, and if both the camera lenses and the films used have the same resolution, the consumer system may have higher resolution than the data reported here. By comparison, 65mm motion picture is 3.6 times the area of a S35 1.85 frame and so has higher resolution capturing even more information from the scene.

If you would like a copy of the paper that further describes this research e-mail roger.morton@kodak.com and ask for the paper described on the Kodak website.