Photo CD

Photo CD images are stored as a hierarchy of components ranging from Base/16 to 64Base which relate to the varying image resolutions that can be reconstructed.[27] The Base image, which has a luma resolution of 512 lines by 768 pixels, is typically used for the TV systems. The higher-resolution images are usually used for photographic imaging. The Base and lower resolution images are usually used for file indexes and thumbnails. The components used to create the Base/16 through 16Base resolutions are stored in a single file called an Image Pac. The Base, Base/4, and Base/16 images are stored uncompressed in order to allow easy extraction for display. 4Base and 16Base images are stored compressed, and as deltas (incremental images) from the resolution below them. Note all resolutions up the maximum resolution in a file (actually an Image Pac) are all present simultaneously. So, for example, a 16Base file contains a Base/16, Base/4, Base and 4Base image as well as the 16Base components. The sixth component, 64Base, is stored in separate files on the Photo CD as part of the IMAGE PAC Extension (IPE). This only exists on Photo CD Pro Master discs.

PCD images use a gamma transformed PhotoYCC encoding.[27][28] Under this scheme, in order to encode scanned images into Photo CD file, the first step is to preshape the RGB data via a non-linear transform (the Rec. 709 OETF flipped to become an odd function, same as the later xvYCC[29]):

${\displaystyle {\mathit {RGB^{\prime }={\begin{cases}\alpha \cdot RGB^{0.45}-(1-\alpha ),&RGB\geq \beta \\-\alpha \cdot (-RGB)^{0.45}+(1-\alpha ),&RGB\leq -\beta \\4.5\cdot RGB&-\beta <RGB<\beta \\\end{cases}}}}}$

In Kodak documentation, α = 1.099, β = 0.018. However, the full standard values (1.099296826809442, 0.018053968510807) provide mathematical continuity.[30]

As a result of this function, colors that are outside of the gamut defined by the CCIR 709[lower-alpha 1] primaries are encoded by the negative values.[27] Linear RGB values are from -0.20 to 2.00, while nonlinear RGB' are from -0.43357 to 1.402278. Reference white (perfect, non-fluorescent, white-reflecting diffuser in the original scene) is 1.0, just like in BT.709.

The preshaped RGB values are then converted to a luminance and two chrominance components via a CCIR 601-1-like matrix:[27]

${\displaystyle Y^{\prime }=0.299R^{\prime }+0.587G^{\prime }+0.114B^{\prime }}$

${\displaystyle B^{\prime }-Y^{\prime }=-0.299\cdot R^{\prime }-0.587\cdot G^{\prime }+0.886\cdot B^{\prime }}$

${\displaystyle R^{\prime }-Y^{\prime }=0.701\cdot R^{\prime }-0.587\cdot G^{\prime }-0.114\cdot B^{\prime }}$

Finally the luminance and chrominance components are scaled to 8-bit values by the following equations:[27]

${\displaystyle Y^{\prime \prime }={255/1.402}\cdot Y^{\prime }}$

${\displaystyle C_{1}=111.40\cdot (B^{\prime }-Y^{\prime })+156}$

${\displaystyle C_{2}=135.64\cdot (R^{\prime }-Y^{\prime })+137}$

Kodak claims that "the scale factors and offsets for the C1 and C2 channels result from the distribution of real world colors".[27] Given the 8-bit integer range for each of the 3 components, 0 ≤ Y ≤ 1.402, -1.40036 ≤ B' - Y' ≤ 0.888689, -1.01003 ≤ G' - Y' ≤ 0.86995.

An oddity of this encoding scheme is that it allows Photo CD images to represent colors which are above 100% (up to 140.2%), "whiter than white".[31] At the time of Photo CD's introduction, this allowed the display of images on the analog televisions of the day to show brighter, more vivid colors than would otherwise be the case. This was acceptable as the analog televisions of the day were designed to be driven beyond their 100% level without abruptly clipping. However, this aspect of Photo CD encoding is not compatible with modern image formats, and is the cause of the well documented "Blown Highlights" problem when converting Photo CD images to more modern formats.[32][33] All modern photo formats enforce hard clipping at 100% (video formats in YUV still have headroom that is present on high level TVs), resulting in highlight clipping in converted images unless some form of Photo CD specific luminance and chrominance mapping is performed.[32]

In addition, although "toeroom" (Y < 0) is not provided, Kodak expressly allows the use of R, G, B < 0, allowing out-of-gamut (for Rec. 709) colors to be expressed.[27] The effect is similar to xvYCC, which came much later.

Photo CD images use three forms of compression in order to reduce image storage requirements.[27] Firstly, chroma subsampling reduces the size of the images by approximately 50%. This subsampling is by a factor of 4 for 4Base images, and a factor of 2 (4:2:0) for all other resolutions. Secondly an additional reduction in size is achieved by decomposing the highest-resolution image data, and storing the 4Base, 16Base and 64Base components as residuals (differences from pixels at the previous level of resolution). Thirdly and finally, the Photo CD system employs a form of quantization and Huffman coding to further compress this residual data. This Huffman encoding is performed on an image-row-by-image-row basis. The Huffman tables are encoded into the Photo CD image itself, and have different lengths depending on the compression class. These Huffman classes are:[34]

• class 1 - 35mm film; pictorial hard copy,
• class 2 - large format film,
• class 3 - text and graphics, high resolution,
• class 4 - text and graphics, high dynamic range.

For example, to recompose an image of 1024 lines by 1536 pixels, the 512-line by 768-pixel luma Base image (which is neither "residual" or Huffman compressed) is interpolated to a 1024-line by 1536-pixel image. A 1024-line by 1536-pixel 4Base residual is then decompressed from its Huffman-encoded form, and its elements are added to each corresponding pixel. The resulting image contains detail for the full 1024-line by 1536-pixel sampled image. To recompose an image to a resolution of 2048 lines by 3072 pixels, the process is essentially repeated, using both the 4Base and 16Base residual. A similar operation is subsequently used to recompose the 4096 line by 6144 pixel resolution. In each step, an identical process is applied to the chroma channels.

Kodak based the PhotoYCC scheme on some existing video standards, notably on aspects of CCIR Recommendation 601 used with PAL and NTSC digital television systems, and also on CCIR Recommendation 709 (now ITU-R Recommendation BT.709) used for HDTV. The PhotoYCC definition is defined in a manner that is not constrained by the limitations of the actual video display.[27][31]

In practice the color space of Photo CD images varies significantly from Rec. 709. Firstly, the Photo CD encoding scheme allows greater than 100% values for color components, thus allowing Photo CD images to display colors outside of the nominal Rec. 709 gamut.[32] In addition, in order to achieve accurate color reproduction, especially when scanning reversal film, Kodak found it necessary to provide ICC color profiles specific to film type and scanner.[35] As a result, by the time that the Photo CD format fell into disuse, five different color spaces were in common use in Photo CD images (PCD 4050 is a Kodak scanner model number):

• Color Negative
• Universal E-6
• Universal K-14
• PCD 4050 E-6
• PCD 4050 K-14

Thus, while the use of the generic Photo CD color space will provide color reproduction that is acceptable for many purposes e.g., Web viewing of thumbnail images, in order to achieve fully accurate color reproduction, e.g., for photographic purposes, any Photo CD display or conversion software must use the correct color profile for the combination of the original media and scanner model.

The table below compares some of the better known options for converting Photo CD images.[38][39][40]