The following are some random Matlab/OpenCV files used in UCF's TRECVID High Level feature extraction submission. The OpenCV files are not compiled and require that you have OpenCV installed. All files come with absolutely no guarantee!!
- Lukas Kanade With Pyramids Dense Optical Flow-OpenCV code
- Usage: ./OpticalFlowDenseLK image1 image2
Computes the Optical flow between images 1 and 2 and then saves the information x_component.xml, y_component.xml, and magnitude.xml.
- Read XML File-Matlab Code
- Usage matrix=read_file('file_path');
Reads in the xml file produced by OpticalFlowDenseLK and returns a matrix with the optical flow information.
- Modified LIBSVM with built in histogram intersection kernel
struct=svmtrain(train_labels, train_points, '-t 5');
[predict, accuracy, probabilities]=svmpredict(test_labels, test_points, struct);
- Keyframe Extraction Algorithm
- This algorithm was based on the paper "ONLINE, SIMULTANEOUS SHOT BOUNDARY DETECTION
AND KEY FRAME EXTRACTION FOR SPORTS VIDEOS USING RANK TRACING" by Wael Abd-Almageed
(University of Maryland). It also uses the function mmread by Micah Richert and the color
space converter by Pascal Getreuer.
**This was NOT the algorithm that was used in our project. It was another one I worked on before another group member wrote a different algorithm...it is thus not completely refined.