The Unofficial Colour Recovery Wiki

The input file containing source material for Colour Recovery should accord with the following specification:

1. Sampling structure[]

1.1 The Colour Recovery process uses input frames sampled at 1920 pixels per line and 1080 lines per frame (2073600 pixels per frame in total). The frame rate is assumed to be 25 fps.

1.2 The film frames are scanned left-to-right and top-to-bottom, as per a conventional (progressive) TV raster, i.e. '1080p'.

1.3 The input frames are assumed to correspond to the full active picture area on the film, i.e. approximately equivalent to the original 4:3 raster fed to the Film Recorder. The pixels are not square.

1.4 It is preferable for the scanned area to be slightly larger than the picture rather than for any cropping to take place.

2. Data structure[]

2.1 The data consists of an 8-bit, unsigned luminance value (0-255) for each pixel.

2.2 The Colour Recovery process assumes that black level corresponds to approximately value 16 and that peak white corresponds to approximately value 235. Sub-black and super-white pixels should not be clipped.

2.3 The gamma is assumed to correspond with that of standard video luminance (i.e. approximately 2.35). Note that the value 2.8 quoted in the 'White Book' is widely accepted to be incorrect.

3. File structure[]

3.1 A CR source file contains raw, uncompressed, luminance data with no header or metadata. Each frame consists of exactly 2073600 bytes of data.

3.2 Multiple frames may be stored within a file, by concatenating the data in temporal order. Hence the size of a file containing N frames (N/25 seconds duration) is N*2073600 bytes.

4. Chunk structure[]

4.1 For compatibility with some file systems (e.g. FAT32) and to keep processing times manageable, each source file should contain no more than 2 Gbytes of data, corresponding to a maximum of 1035 frames (approximately 41 seconds).

4.2 Sequences longer than 41 seconds should be split into 'chunks' each contained in a separate file. It is recommended that the size of each chunk (except the last) should be 1000 frames, or 40 seconds.

4.3 Chunks may //optionally// have an overlap to reduce the impact of 'end-effects'. For example, each file may consist of 1025 frames, where the last 25 frames of each file contain the same data as the first 25 frames of the next file. However, it has not in practice been found essential to do this, and for simplicity it is recommended that there be no overlaps.

5. File naming[]

5.1 The source files should be given the extension '.y', indicating raw luminance data.

5.2 It is recommended that each file is named to indicate the programme material it contains and the chunk number. For example a file containing chunk 11 of episode 2 of The Mind of Evil might be named 'moe_ep2_ch11.y'.

5.3 It is recommended that chunk numbering starts at zero, i.e. chunk 00 corresponds to the first 1000 frames, chunk 01 to the next 1000 frames and so on. A 30-minute programme would require 45 chunks (00-44).

5.4 For convenience, and to improve performance with some file systems, it is recommended that the chunks corresponding to each episode of a series or serial be contained in a different directory (folder).

Version 1.0, 28-May-2009, Richard Russell.

To save a file in the appropriate format from Adobe® Photoshop®[]

  1. Convert the image to monochrome. From the Image menu select Mode > Grayscale. While you have the Mode menu visible, check that "8 bits/channel" is ticked.
  2. Resize the image if necessary. From the Image menu select Image Size. Enter the Width (1920) and Height (1080) values, unchecking the "Constrain Proportions" box if necessary. Make sure the "Resample Image" checkbox is on, and choose an appropriate resampling algorithm from the dropdown; this may need some experimentation.
  3. Save the image. From the File menu select Save As..., choose "Photoshop RAW" from the dropdown list and replace the ".raw" file extension in the filename with ".y".