Conceptual Software Produces Amazing Detail From Bad Lenses
The standard motto regarding photography equipment is “invest in glass” with good reason. While sensor technology certainly has a tremendous impact on the quality of the photograph, the real limitation often lies in the optics. One need only look at the difference between a kit lens and a top-level prime to prove this. However, new methods of image processing may soon alter the effectiveness of sub par glass. Researchers at the University of British Columbia have made some serious headway in a project they call “High-Quality Computational Imaging Through Simple Lenses.” The project in it’s current form uses simple lenses, meaning a lens with a single glass lens element rather than the complex multitude of optical elements in a standard camera lens. This means that these test lenses suffer from problems that complex lens systems have either lessened or completely eliminated such as chromatic aberrations, geometric distortion, and field-curvature to name a few.
Fixing these problems post-image capture is something that even highly advanced raw processing algorithms still struggle to perfect from amazing lenses. This team is showing that it can be done effectively using what equates to cave-man photo gear.
The result speak for themselves and are obviously pretty exciting when looking into the future for what this could be developed into. If this can be done with horrible lenses, imagine what could be done with just average optics.
Turns out they’ve started to answer that as well. The images below were shot with a Canon 28-105:
If you ask me that’s pretty incredible given that it’s still very much a work in progress. As you probably heard in the video there are still quite a few limitations. Even still, I would imagine that a certain software company will be snatching these developers up very soon.