Image Manipulation Experiments

[sambux1]

We need a way of determining which parameters to use for the perceptual hashing layer in the perceptographic hash function. To do so, we need to be able to manipulate images to see how the perceptual hashes change for each of the perceptual hash functions we support.

For each type of manipulation, we want to be able to manipulate some image by some quantifiable amount and see how it affects the perceptual hashes. That is, we want to have a function that takes as input an image and some "manipulation factor" and output a modified image.

For example, one type of manipulation is compression, so we want to be able to compress images by some factor (e.g. 10%, 50%...).

The first step of this is to simply define one function per type of manipulation to actually manipulate an image. Each should have a signature of the form

manipulation(image, factor) -> modified_image

I'm assigning this step to @ahmedh409. After this step, we can work together on writing a class to use the manipulation functions to actually get data about the perceptual hash functions and their parameters.

An incomplete list of the types of manipulations we want are below. Feel free to add to this.

• compression
• gaussian blurring
• white noise