recoverjpeg(1) recover jpeg pictures from a filesystem image


recoverjpeg [options] device


Recoverjpeg tries to identify jpeg pictures from a filesystem image. To achieve this goal, it scans the filesystem image and looks for a jpeg structure at blocks starting at 512 bytes boundaries.

Salvaged jpeg pictures are stored by default under the name imageXXXXX.jpg where XXXXX is a five digit number starting at zero. If there are more than 100,000 recovered pictures, recoverjpeg will start using six figures numbers and more as soon as needed, but the 100,000 first ones will use a five figures number. Options -f and -i can override this behaviour.

recoverjpeg stores the recovered pictures into the current directory. If you want it to store them elsewhere, just go to the directory you want recoverjpeg to save the images into (using the cd command at the shell prompt) and start recoverjpeg from there, or use the -o option.

Note that device is not necessarily a physical device. It may also be a file containing a copy of the faulty device in order to reduce the actual processing time and the stress imposed to an already defective hardware. dd(1) or ddrescue(1) may be used to create such a working copy.


Display an help message.
-b blocksize
Set the size of blocks in bytes. On most file systems, setting it to 512 (the default) will work fine as any large file will be stored on 512 bytes boundaries. Setting it to 1 maximize the chances of finding very small files if the filesystems aggregates them (UFS for example) at the expense of a much longer running time.
-d formatstring
Set the directory format string (printf-style, default: use the current directory). When used, 0 will be used for the 100 first images, 1 for the 100 next images, and so on. The goal of this option is to circumvent the directory size limit imposed by some file systems.
-f formatstring
Set the file name format string (printf-style, default: "image%05d.jpg"). It is used with the image index as an integer argument.
-i integerindex
Set the initial index value for image numbering (default: 0).
-m maxsize
Maximum size of extract jpeg files. If a file would be larger than that, it is discarded. The default is 6 MiB.
-o directory
Change the working directory before restoring files. Use this option to restore files into a directory with enough space instead of the current directory.
Be quiet and do not display anything.
-r readsize
Set the readsize in bytes. By default, this is 128 MiB. Using a large readsize reduces the number of system calls but consumes more memory. The readsize will automatically be adjusted to be a multiple of the system page size. It must be greater than the maxsize parameter.
-s cutoffsize
Set the cutoff size in bytes. Files smaller than that will be ignored.
-S skipsize
Set the number of bytes to skip at the beginning of the filesystem image. This can be used to resume an interrupted session, in conjunction with -i. The number of bytes may be rounded down to be a multiple of a memory page size in order to improve performances.
Be verbose and describes the process of jpeg identification. By default, if this flag is not used, recoverjpeg will print a progress bar showing how much it has analyzed already and how many jpeg pictures have been recovered.
Display program version and exit.

All the sizes may be suffixed by a k, m, g, or t letter to indicate KiB, MiB, GiB, or TiB. For example, 6m correspond to 6 MiB (6291456 bytes).


Recover as many pictures as possible from the memory card located in /dev/sdc:

recoverjpeg /dev/sdc

Do the same thing but ignore files smaller than one megabyte:

recoverjpeg -s 1m /dev/sdc

Recover as many pictures as possible from a crashed ReiserFS file system (which does not necessarily store pictures at block boundaries) in /dev/sdb1:

recoverjpeg -b 1 /dev/sdb1

Do the same thing in a memory constrained environment where no more than 16MB of RAM can be used for the operation:

recoverjpeg -b 1 -r 16m /dev/sdb1


Copyright (c) 2004-2013 Samuel Tardieu <[email protected]>. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

If recoverjpeg saves your day and you liked it, you are welcome to send me the best rescued ones by email (please send only 800x600 versions of the pictures) and authorize me to put them online (indicate which contact information you want me to use for credits).


Recoverjpeg does not include a complete jpeg parser. You may need to use sort-pictures afterwards to identify bogus pictures. Some pictures may be corrupted but have a correct structure; in this case, the image may be garbled. There is no automated way to detect those pictures with a 100% success rate.


Samuel Tardieu <[email protected]>.