The bilateral filter was originally introduced as a simple, non-iterative technique for edge-preserving smoothing of images. The bilateral filter has found widespread use in several image processing, computer graphics, and computer vision applications such as image denoising, video abstraction, demosaicing, optical-flow estimation, and stereo matching.

The direct implementation of the bilateral filter, however, is computationally intensive. Very recently, we discovered a simple algorithm for implementing the filter in real time. The highlight of our algorithm is that allows us to process the image using a fixed number of operations per pixel, independent of the size of the image and that of the kernel. Moreover, our algorithm can be implemented in parallel, and this allows us to further accelerate the speed of the final implementation. The final output is comparable and, at times, even superior to the state-of-the art implementations of the bilateral filter.

In this project, the student will develop novel applications for processing, and subsequently analyzing, biological images using the fast bilateral filter. The student will be provided with the Java implementation of the filter.