Shyam Naren Kandala
Anjaneya Srujan Narkedamalli
Introduction:
Seam carving is a method for image resizing without distorting the essential features of an image
used for displaying an image on devices with heterogenous screen sizes. This is done by assigning each
pixel an energy value (which signifies its relative importance of a pixel in an image),
and identifying either horizontal or vertical seams of pixels with least energy sum for
removal to reduce the image size.
Proposed Work:
The aim of the project is to parallelize two different Seam Carving approaches and
analyze their performances on UCI HPC GPU cluster. All the approaches start by the computing
the energy function using ‘‘Gradient Magnitude’ which is embarrassingly parallel . Then they
identify the seams to be removed, parallely, as follows:
- Compute the minimum energy sum path using Dynamic Programming.
- Compute the minimum energy sum path using Greedy algorithm (Dijkstra’s)
We will also implement a serial version of seam carving using dynamic programming to use it as test oracle.