Scanning resolution is one key issue when it comes to scanner performance. Many tests and demonstrations have been done to show the ultimate resolution of film. Most scanning is currently carried out at 2K resolution, i.e. the image is sampled with 2048 pixels per line. For comparison, HDTV has 1920 pixels per line. 4K (4096 pixels per line) scans are sometimes performed for VFX, and some productions, Spider Man 2 for example, have been done entirely in this format. Looking at the digital files in Figure 3.1, a 4K scan definitely contains more details than a 2K scan. The trade-offs are scanning time per frame, data transfer times, and processing costs.

Care needs to taken that no aliasing appears in the scans. In digital imaging, aliasing is a term used when patterns are created in an image that were not in the original. It can be avoided by resampling to a lower spatial resolution with a digital filter or by an optical filter. Optical anti-aliasing filter, however, tend to produce less sharp images.

Dynamic range is the other key issue, meaning the scanner needs to capture the entire density range of the negative. Even though this may contain information that will not be shown in the final image, it gives headroom to change the dynamics of a shot to fit creative requirements.

Together with dynamic range comes the question of quantization. A scanner has to translate the continuous range of densities in a negative into discrete digital numbers. The bit depth or number of bits per channel determines how many levels can be encoded. Usage of 8 bits results in 256 levels, a number generally considered too low for negative film scanning. One needs more than 100 levels to produce a perfectly smooth tonal scale. Lower quantization breaks graduated colors into visible blocks. The effect is called banding or contouring and is illustrated in Figure 3.3 and Figure 3.4. Color processing may make banding visible where the tonal variation of the original image appears to be smooth. Therefore, scans with a higher bit depth than 8 bits are needed as source for the color correction.

A commonly used file format for film scans is the Cineon or DPX format. It uses 10 bits per channel, which equals 1024 levels, and linearly encodes the densities of the negative above base as shown in Figure 3.5. One code value in the DPX file represents 1/500 log D. A look at Table 3.1 reveals that the base density of the negative is encoded as 95 rather than zero. That happens for the same reason an IN is about 0.2 log D heavier than the original negative (see the section called “Characteristic Curve”). Recording the Cineon image on intermediate film results in a digital IN, where the densities are shifted by 0.19 log D. More information about the Cineon and DPX format is given in the appendix.

Figure 3.6 displays the sample image in log encoding. Figure 3.7 shows 5 enlarged portions of the image together with the code values from the scan of the negative. A DPX file presents a positive image but otherwise it keeps the characteristic of a negative. One could regard it as the digital version of an interpositive. It looks “flat”, the blacks are too high, and the whites are to low since the additional tonal values below and above are linearly encoded as well. Like film negatives, Cineon/DPX images are not meant to be judged by the human eye.

Nevertheless, a scanner operator or colorist has to examine the scans. Since a film scanner applies no color correction, the scan may not look neutral. The midtones may not necessarily sit around a code value of 470 because some cinematographers prefer to overexpose the negative by one or two stops. Examine the histogram in Figure 3.8. It shows the relative frequency of the code values in the sample image. The number of pixels towards the lower and upper limit of the range is zero. A DPX image with a significant number of pixels having a value of 0 or 1023 is most likely the result of a misalignment of the scanner. Also, there should be no gaps in the histogram. Natural images always have a continuous variation of tonal values. Remember to display the histogram 1024 pixels wide to recognize any gaps. Figure 3.9 gives an example of a histogram showing missing code values in the higher densities.

Each line in the table corresponds to an exposure increment of one third stop. The numbers in the column titled “Label“ refer to the markers in Figure 1.1. Enlarged portions of the marked areas are displayed in Figure 3.7.

Most film scanners use a CCD (charge coupled device) or a CMOS (complementary metal oxide semiconductor) element as image sensor. Both types of imagers convert light into electric charge and process it into electronic signals. The response of both sensors is related to the transmittance of the film. However, the important photographic quantity is the density.

Although one can easily convert transmittance to density, there is a difficulty. The conversion function is very steep at high opacity/density values, as shown in Figure 3.11. Therefore, more code values or more bits are needed in the sensor output than in the final file format, the film transmittance must be sampled with a higher quantization than the density.

Some people prefer to work with linearly encoded images instead of the Cineon/DPX format. Confusingly, there are at least three different types of encoding referred to as “linear”.

Although it is not regarded as “linear” the DPX format is actually a linear representation of log scene luminance, compare Figure 3.13 with Figure 3.14.