I guess it will depend on how the files are named. If you have only ever used the default file names, and have effectivly only changed the extention when converting from RAW to JPEG, and the JPEG only images are not from Canon camers you should not have to much difficulty. In your file browser of choice simply view by extention, this will then group all the JPEG images together, and usually sub sort them alphabetically by file name. You then just need to find the images that start img_ (or _img_ if you shot with AdobeRGB as your in camera colour space) and delete them. I would though first check the name/number ranges of the RAW files to ensure that I only got corresponding files. This could also be automated if you have any batch file/scripting experiance
If you have changed the filenames at any point in this process so that the names of the JPEG's and the names of the RAW's nolonger match, and there is no simple way of differentiating the images you wish to keep from those you want to delete it will be much harder. If your file browser can sort on image EXIF data, and you did not strip it from the JPEG files, you could sort on the date/time of the exposure. Windows Explorer will do this, and at least the RAW and JPEG images should at least be next to each other now, but you will have to select every image individually. This could still be programed but will be harder as you would have to be able to read the EXIF data in both files. You would also have to check the EXIF of every single JPEG file for each RAW. If you know that there is only one JPEG for each RAW then you could have it drop out of that loop after it finds a match, which would speed things up a bit.
I have written DOS batch files in the past that did the first type of sort/delete a long time ago, Pearl/PHP would probably be the best option now, but couldn't now it. The second option would need at least Pearl/PHP and may even need something more powerful, as you need to be able to read/match the EXIF data. If you do not have the EXIF data in the JPEG files then you are stuck doing a visual comparison of the files, and I know of no reliable way to do that, especially between different file formats, as a bit for bit comparison of the files is not an option. At least to JPEG files that are say 90% identical in a bit for bit comparison may well represent two images that are the same except for minor edits or differences in compression, although the differences, even just from two different runs through the compression may be very significant, I do not know as it is not something I have thought about much.
Alan
