Aim:
The purpose of this assignment is to develop a program to segment a video using a shot-boundary detection technique and then extract key-frames.
Facilities and Basic Programs:
You can use the code provided for Assignment 1 but it is not necessary. Three test video sequences (each as a mpeg clip and as a set of jpeg frame images) have been made available to you. They reside on:
Deliverables:
You are required to write a program to perform shot-boundary detection using the twin-comparison approach. The program should take a mpeg video clip (or a set of jpeg frame images) as input and output the frame numbers denoting the shot boundaries (both abrupt as well as gradual transitions). You need to determine the necessary thresholds appropriately. Note that values given in the paper may not necessarily work -- some experimentation will be required.
Taking the shot information obtained, you need to extract key frames which represent the shots using the average histogram technique. You program should be flexible enough to allow for the extraction of n key-frames per shot where 0 < n < 10 is an input parameter.
Discuss the performance of the twin-comparison technique on the test video sequences. Explain the reason for the algorithms' success/failure on these sequences. Also comment on the quality of the key-frames obtained for several values of n.
Deadline:
The complete programming and the report should be ready for evaluation by 12:00noon on 15th April 2002. You are required to:
Grading Scheme: