Images can just as easily break laws, send out abusive immoral messages, offend staff and damage company reputation. Images coming from your company can make you liable to prosecution and attract unwanted media. Some firewall proxy servers support image scanning usually purchased as an additional subscription, which can detect pornography and other unsuitable material.
Image Manager from Clearswift's MIMEsweeper for SMTP not only scans for images within the body of an email, but also scans for images deep within attached files, completely breaking them down extracting embedded images and passing them on to Image Manager to scan for unwanted images.
When Image Manager is scanning for images, first it checks its pre-populated database. It checks the hash of the image and compares this hash with hashes within it's database. These hashes within the database are a bulk of known images downloaded from Clearswift servers periodically.
If there is no match within the database then image manager scans the image and looks for certain characteristics within the image to classify appropriately. These features and characteristics are compared to it's built in image feature database. This database contains values for thousands of features, and if the feature matches with one in the database or even a close match, then appropriate action is taken depending on the classification of the matched feature within the database. So if the feature that it is matched to within the database is known to be a violent image, then image manager will classify this scanned image as spam.
Wikipedia's guide to Spam Techniques