Is it possible to do very basic image recognition to compare an image against a database of images and determine which image in the database is the best match? I don't need to do any processing of any of the images, but simply differentiate between a finite list of images.
Are there any libraries that make this somewhat possible? Or would it require building an image recognizer from scratch?
Take a look at OpenCV. OpenCV is not optimized for iOS but it can be a good start. If I understand what your are doing is it not really easy to achive and I did not know if you can find opensource library other than OpenCV.
I am still not getting exactly what you want to achieve.
Do you want to i.e. compare the different smiley images you get with the happy or sad reference smiley picture and see whether one of the given references is shown in that picture or do you really want to extract whether the smiley/person on that image is smiling?
Latter is not so easy to achieve. Most attempts to get emotions of people on pictures extract the different features there can be seen in the face, i.e. width of mouth, position and "openness" of eyes, position and angle of eyebrows and so on... you would need to implement methods to get these features (there are several methods for that) and classify the feature-vector to your given test results to get a best match on which emotion it is... But all this is really not done easily...
If you really would like to compare how similar two images are or whether one image is in the other (i.e. whether there is that yellow smiley in an image), you could try a template match on the two images...
