In the search for exoplanets, many methods have proven fruitful. All involve careful observation of the host star and complex post-processing algorithms to identify if any planets are part of the system. One methodology that has shown promise, yet currently yields relatively few results, is direct imaging. The first part of this dissertation showcases the development of a novel technique to identify planets in post-processing of direct imaging data. It leverages the common spatial pattern filtering algorithm in combination with a forward modeled matched filter. I compare the algorithm to other leading techniques. The second part develops the tools and software for generalizing this approach to many different datasets. This allows for systematic, large-scale statistical analyses of the CSP method applied to a variety of stars and injected data. I present results for multiple sets of observations and show how the new technique can be expected to perform. Finally other avenues of image processing are explored both for use in a new type of filter and for the development of advanced self-assembling telescopes in space.