Motion estimation includes the process of determining the motion vectors of different areas in the frame being encoded. Typically, motion estimation is very computationally demanding, so it is the bottleneck in MPEG video encoding.
- Find similarities between video frames and exploit this to achieve high compression rates
- Code only the relative movement of current frame to reference frame
- Use Sum of Absolute Differences (SAD) as the measurement metric
- SAD typically computed for 16x16 blocks
Our program will take in two images, one being the current and the other a reference image, and will output the motion vector and differences in the images. It will also output the minimum SAD value.
Example current and reference images:
In order to run the application, just run the following commands:
-
To generate images using our image generator:
python3 images/image_generator.pyand may need to runpip3 install pillowbefore this as well -
To create the executable file:
gcc -o <path_to_exec_file> motion-detection.c -lpng -
To run the executable file:
.<path_to_exec_file> <path_to_current_image> <path_to_reference_image>
- Isaac Sahle
- Shreyas Devalapurkar