Description
Description: In this project, you will• experiment with video data representation,• feature extraction, and• experiment with video retrieval.Tasks:The input to the project will be a video file.You will operate in the Y color space.• Task I: Implement a program which divides the frames of the given video into 8-by-8 regions – called frameblocks. Then, given a user supplied n, the program performs the following subtasks:– Task I(a): The program creates a n-bin gray scale histogram, by quantizing the Y color space, for eachcellblock of each frame in the given video. The outputs, of the formhframe id, block coord, gray instance id, num pixelsiare written into a file of the formvideo filename hist n.hst– Task I(b): The program applies 2D-DCT on the Y component of each block of each frame in the givenvideo and selects the n most significant frequency components. The outputs, of the formhframe id, block coord, freq comp id, valueiare written into a file of the formvideo filename blockdct n.bct– Task I(c): The program applies 2D-DWT on the Y component of each block of each frame in the givenvideo and selects the n most significant wavelet components. The outputs, of the formhframe id, block coord,wavelet comp id, valueiare written into a file of the formvideo filename blockdwt n.bwt– Task I(d): For each frame (except the last) of the video and for each block, the program computesthe difference from the same block in the next frame and the creates an n-bin difference histogram, byquantizing the obtained differences. The outputs, of the formhframe id, block coord, diff comp id, pixelcountiare written into a file of the formvideo filename diff n.dhc• Task II: Implement a program which applies 2D-DWT on the Y component of each image video frame in thegiven set and selects the m most significant wavelet components (m is user supplied). The outputs, of the formhframe id,wavelet comp id, valueiare written into a file of the formvideo filename framedwt m.fwt• Task III Implement a program which given (a) a query video, (b) a frame id, and (c) n and m values, identifiesthe best 10 matching frames (except itself) for each of the 5 feature types and displays the original frame aswell as the 10 matches (and matching scores) for each feature type.Deliverables:• Your code (properly commented) and a README file.• A report describing your work and the results on a sample image.Please place your code in a directory titled “Code”, the outputs to a directory called “Outputs”, and your report ina directory called “Report”; zip or tar all off them together and submit it through the Blackboard.
