Image Super Resolution: Segmentation Results

The hardware used to run the segementation is a i7-5600U Processor (due to the lack of a dGPU in this device, Thinkpad T450s).  Using a more powerful processor will improve the runtime of the algorithm.  However, as we will see the runtime of the segmentation algorithm (CPU implementation) is insignificant to the runtime of the upscaling therefore there is not much to be improved from a hardware point of view.  Note: Other than the first test image, it appears that a sigma of 0.8 (for smoothing seems to work best.

Here is the source image of an image to segment.  It is composed of 3 colors, a black background and 2 colored bars.

1

These calculations were run with the input arguments k=30, min_size=10, sigma=0.1 and had a run time of .153 seconds for 3 segments.

This slideshow requires JavaScript.

For the next image the calculation and output of the segments took 9.666 seconds and created 17 segments.  In lieu of space, only the most significant segments will be shown.

1

Here is the source image.

This slideshow requires JavaScript.

As we increase the complexity of the image, we will notice that the number of segments to begin growing exponentially.  Now we look at the final example of image segmentation.

1

Here is the source image.

The input arguments were sigma=0.8, k=750, min_size=2000, had a run time of 10.320 seconds and produced 27 segments. The input arguments were sigma=0.8, k=750, min_size=2000, had a run time of 10.320 seconds and produced 27 segments.

This slideshow requires JavaScript.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s