SkyReplacement

Final Project for Computer Vision

Group members: Josh Stern, Savvas Petridis, Eli Fessler

Objective:

Take in a sequence of JPEG images and replace a high contrast sky with another texture.

Components

1. Parallelization of necessary algorithms
2. Frame to Frame data structure
3. Sky isolation
4. Texture application

Algorithm Flow

1. Threshold each image as to isolate the sky (allowing us to do multiplication for replacement)
2. Track points from one original image to the next
3. Using these points, calculate homography matrix
4. Use Homography to calculate where to sample from sky texture image (so it moves appropriately)

Input/Output Details

One image:

./imgpro <full_input_image_name> <full_output_image_name> 1 <-flag stuff>

More than one:

./imgpro <general_input_image_name> <general_output_image_name> <image_count> <-flag stuff>

This will append "_i.jpg" to all general names, where i is an integer (starting from zero).

Unexpected Results

If images are flipped (due to extra image metadata), install Jhead using brew install jhead (Homebrew required) and run

jhead -autorot path/to/image.jpg