Is it easy to tell if digital images are corrupted? Do corrupted images always show obvious signs of corruption, such as missing parts of the image? Or could corruption be more subtle, perhaps just causing slight color changes?
In other words, if an image doesn't look obviously corrupted, can I assume that it is not corrupted and that the colors in the image are accurate?
No. It's possible for a bit to be flipped which has no immediately obvious effect. In fact, it's possible for a data error to have no visual effect at all.
If reasonably-sized blocks are missing or scrambled in the middle of the file, you'll get the kind of visual artifacts we see in typical questions under data-corruption. If a file is just truncated at the end, you may see just a black line at the bottom — easy to overlook. For JPEG files, either of these types of problems are usually detectable as the file format will not match what's expect from the spec, and tools and methods at Is there a tool to check the file integrity of a series of images? will usually detect this.
Even a single bit flip can be significant, since most graphics file formats have compression without any error correction. That usually means that everything after that point is confused (and visibly ruined).
On the other hand, there are a lot of areas of most files that can change without making any difference you can spot by eye or by a tool unless you have a checksum of the original to compare. This is also the case if the error happens to be in the metadata — maybe you'll loose some EXIF information, or have scrambled XMP tag data, but still have a valid file.
Most of the time, yes, but you can't count on it.
A single flipped bit in a image file will usually make a mess due to the image encoding. It's rare that images are just verbatim dumps of pixel data, since even lossless compression yields reasonable savings most of the time.
Changing one bit in something that is LZW compressed (a lossless compression common in TIFF files), for example, makes a mess because it doesn't just affect one bit in the uncompressed pixel data. It will usually result in at least a horizontal streak of something obviously wrong, and can often confuse the compression decoder to the point that the whole rest of the picture is gibberish.
