Platform independent tools for working with Lytro LFP files

Lytro has not announced much about their file formats other than that they have software for OS X and will have Windows supported in 2012. To enable support for other platforms, it will be useful to develop open source software to process their files.

This tool supports both the large raw files that come from the Lytro camera and the compressed files that the desktop software produces for web display.

Note: The description below refers to .lfp file format for files generated using Lytro's Version 1 processing software. See README_V2 for a description of .lfp format changes as of Lytro's December 2012 update (Version 2).

.lfp file format

The file itself is formatted as follows. First, a header:

# magic 12 byte header (LFP)
89 4C 46 50 0D 0A 1A 0A 00 00 00 01
# 4 byte length (0, since there is nothing in this section)
00 00 00 00

After this are a number of sections. The data in the first is plain text JSON table of contents describing what the rest of the sections in the file contain. The remaining sections can be additional metadata, a depth lookup table, compressed jpg images, or raw sensor data depending on the file. The sections are formatted as follows:

# magic 12 byte header (containing a type like LFM or LFC)
89 4C 46 4D 0D 0A 1A 0A 00 00 00 00
# 4 byte length, not including header, sha1, or null padding
00 00 07 A7
# 45 bytes of sha1 hash as hex in ascii
73 68 61 31 ...
# 35 bytes of null padding
00 00 00 00 ...
# the data of length previously specified
7B 22 70 69 ...
# 0 or more bytes of null padding
00 00 00 00 ...

.lfp web files

The Lytro desktop app exports compressed representations of the light field captured by the Lytro camera to reduce file times and rendering requirements for web display. The .lfp files are simply a set of JPEG files representing the unique visually interesting sections of the light field. That is, a set in which each image shows a different area in focus. It appears to do so dynamically, picking the minimum number of images necessary to show all focusable objects in narrow depths of field.

These images are stored along with their estimated depths and a depth lookup table for the image. This allows for HTML5 and Flash applications in which the user clicks on a region of the image, the value of that region is looked up, and the depth image closest to that value is displayed.

.lfp raw files

The files that come directly from the camera are roughly 16MB, made primarily of one section which is a raw Bayer array of the 3280 x 3280 pixel sensor. There are also two metadata files, one containing detailed information about the format of the captured image, and the other containing serial numbers.

lfpsplitter

lfpsplitter is a commandline tool that reads in a .lfp file and splits it into a plaintext metadata file, a plaintext listing of the depth lookup table, and the component jpgs.

make
    gcc -O3 -Wall    -c -o lfpsplitter.o lfpsplitter.c
    gcc -o lfpsplitter lfpsplitter.o  -O3 -Wall 

./lfpsplitter IMG_0001.lfp
    Saved IMG_0001_table.json
    Saved IMG_0001_imageRef0.raw
    Saved IMG_0001_metadataRef.json
    Saved IMG_0001_privateMetadataRef.json

./lfpsplitter IMG_0001-stk.lfp
    Saved IMG_0001-stk_table.json
    Saved IMG_0001-stk_depth.txt
    Saved IMG_0001-stk_0.jpg
    Saved IMG_0001-stk_1.jpg
    Saved IMG_0001-stk_2.jpg
    Saved IMG_0001-stk_3.jpg