Zef : A Computer Vision Library for Haskell

We present a computer vision library for Haskell and discuss its design and architecture. The library is then used for developing two demo applications. Two case studies analyze these demos and show that Haskell is well-suited for implementing vision algorithms, with significant gains in productivity, readability, correctness, and conciseness. We demonstrate that by keeping a limited set of performance-critical operations in C/C++, the loss in performance is insignificant. To further support this claim, we use our library to implement a real-time video processing application on a mobile platform.