Before deciding which algorithm to use for this classification, you must first understand how you analyze the data to see if/how the information obtained can be used for this purpose. It is not unlikely that even before starting the analysis of the data is necessary to perform a pre-processing of the images, such as resizing, color changing, blurring, etc ...
A field of computational science not often mentioned is the
Data Minning, the covers concepts that can qualitatively evaluate the data sets and to group them into separable categories that can better represent the cluster to which a sample belongs. A useful tool in this subject is the
Weka, which natively has several algorithms commonly used, so that you can assess which one will take in your case.