问题
I have a requirement where in I have to determine whether a photo is corrupted and accordingly tag it as such.
Another thing, I need is to determine if an Image has got wrong extension. What I mean by wrong extension is that sometimes I have come across a photo that has extension of jpg but when I load this photo into IrfanView it reports that the photo is in different format that the extension.
How can I do this in Delphi.
回答1:
I have a requirement where in I have to determine whether a photo is corrupted and accordingly tag it as such.
You can try some things, but with certain file formats (example: BMP, JPEG to some extent) only a human can ultimately decide if the file is OK or corrupted. The simplest test is to simply load the file into a corresponding object (TJpegImage, TPngObject, etc). If you get an exception while loading you've surely got a corrupted file. Unfortunately if no exception is raised you can't really say the file is not corrupted. I've seen corrupted JPEG files that load just fine into a Delphi TImage and can be opened with Windows's Image Viewer, but are obviously corrupted to a human observer. With BMP images it's even clearer: open up a bitmap, overwrite some bytes in the middle of the file and then open it in a viewer. How can any automated system tell those wrongly colored bits in the middle of the bitmap are actually wrong?
Another thing, I need is to determine if an Image has got wrong extension. What I mean by wrong extension is that sometimes I have come across a photo that has extension of jpg but when I load this photo into IrfanView it reports that the photo is in different format that the extension.
How about doing some of the same, trying to load the file into the object that corresponds to it's extension, and if you fail, try opening up with some other formats? This should be easy.
Alternatively you can investigate image headers: Most file formats start with a short signature, a few bytes. You can look up the documentation of all image file formats and find the signature, or you can simply open up an large number of files and look for a pattern in the first 4 bytes. I'd go for this second alternative since finding proper documentation for all image file formats might be a challenge.
回答2:
Since you used the term "requirement", I suspect that you're doing a job for someone, possibly as a contract. So make sure that you nail the requirements before worrying about the code.
IMO, you need to get samples of test cases. As others mentioned, failure to load the file as a particular format will be one test. But what about a .jpg that loads ok, but the bottom third is missing? Or a .jpg that loads ok but has green "static" lines in the middle where an error occurred upstream somewhere (on the camera, photoshop, whatever) but then the processing recovered and resumed? In this case, the .jpg may really have green lines in it. Is that considered "corrupt" or not? This is where you need to be careful, especially if it's a contract job.
回答3:
The only way to check if file is corrupted is to try reading it as it is described in file format, ie. load BMP as BMP with reading BMP header, BMP data etc. There are many web pages that describe graphics file formats. Of course if you transmit files and are afraid that it will be corrupted after transmitting then save such files with some sum like CRC32, or even cryptographic MD5 or SHA1. Then after transmitting check if calculated sum is the same as original.
In Delphi there is unit jpeg and types TJPEGImage and TBitmap. Try loading it with data and check exception. For others formats there are many libraries, just look for required file formats.
To check if file extension is good try reading some first bytes of file and check it with some dictionary of graphics file headers. For example GIF files should start with GIF, BMP files starts with BM, and in JPEG header you will find JFIF. I think unix utility file works this way.
来源：https://stackoverflow.com/questions/5138797/how-to-determine-if-a-photo-is-corrupted