A tool for fixing corrupted JPG photos(repost)

A tool which will accept one or more .jpg files, analize them and identify corrupted ones. The corrupted files will be fixed and saved into a new file. ## Deliverables We are looking for a talanted and creative coder that find a solution to a problem that arise mostly due to problems while copying photos from one place to another. Sometimes, even a single bit change, can cause the entire photo to look like in the samples we have provided. The is some interesting open source project that can be studied. <[login to view URL]> However, this tool only analize the photos and allow limited manual operations to be performed on it, but doesn't provide an automatic solution. We will evaulate the provided solution by running it on several corrupted photos attached to this bid request. Some points to be taken into consideration: 1. The tool should go over a folder / folders and identify corrupted images automatically. 2. The tool should alert if during the process of copying photos, a photo is about to become corrupted (the destination version is different than the original. The destination is corrupted and the original isn't), and prevent the completion of the copy, allowing it to be repeated until the destination is identical to the original. Checksum mechanism can be used for doing so. 3. To address the scenarios where the difference between the old photo and the new corrupted photo is a single bit (or several single bits), the tool should create all possible alterations and then validate the result. 4. the JPG format contains sometime several images (for example: a smaller thumbnail version), which can be used to validate the success of a repair. These are only ideas but it's up to you to invent the solution. For further reading: <[login to view URL]> General requirements: A c++ (MFC or Win32) application compatible with Visual Studio 2008. * * *This broadcast message was sent to all bidders on Thursday Mar 3, 2011 4:58:04 PM: Here are several guidelines to all bidders about possible ways to provide the solution (and the proof of concept). 1. If the header of the .jpg file is corrupted, replace it with a generic header of your own. In the real project, please provide the option of importing the header from another photo taken nearly at the same time to replace corrupted headers. That can be even done automatically by trying to use the header of a photo numbered before or after the damaged one. For example if the damaged photo is DSC0111, try importing the header from DSC0112 or DSC0110. 2. When a part of the original data of a photo is missing, please determine incomplete JPEG scans and replace them using AI merhods, information taken from the thumbnails kept ( you would usually find 2 of them: Thumbnail of 160 x 120 pixel and Thumbnail of 640 x 480 pixel), etc. Try to compare to the next or previous photo by the sequencial number and so on. 3. In some cases, there is a single or multiple bit and byte errors in JPEG scans (color error). When a certain piece of data in a JPEG scan is wrong, decoding will result in erroneous data for this data region itself and for all successive data regions in the JPG file. Because the parts of an image are coded horizontally, there appear the so feared faded lines in wrong color. We need a solution for such problems, using trial and error, or some kind of a learning mechanism.