YUCHONG YE

Implemented a paralleled Gaussian blur algorithm for image smoothing. The different color channels are separated into Structure of Arrays (SoA) for coalesced memory access.

Transform High Dynamic Range (HDR) image into 0-255 RGB range so that it can be displayed properly on monitor. Histogram equalization algorithm is used to compress the range. Three typical parallel operation: reduce, scan, scatter are used to compute the cumulative distribution function (CDF).

I first creating a score for every pixel that tells us how likely it is to be a red eye pixel. Then these values are sorted in ascending order so that we know which pixels to alter to remove the red eye. A parallel Radix sort is used for speed up.