hw4.pdf will have the assignment details, the video and img will be used in the image.
hw4-qoe.docx In this homework, we will practice hands on experience with using video codec to encode video sequences, and compute its quality of experience (QoE). For the dataset provided by the TUT Finland, let us select the first 2 sequences for the test: http://ultravideo.cs.tut.fi/#testsequences Data set preparation: Download the mp4 file format/AVC coded source as input: To reduce the computational complexity, we can reduce the sequence spatial resolution to 640x360, via this command: % pull the first 2 sec of frames out, resize to 640x360 ffmpeg -i HoneyBee_1080p.mp4 -ss 00:00:00.00 -t 00:00:02.00 -s 640x360 -c:v libx264 -qp 2 HoneyBee_360p.mp4 % generate 420p YUV output: ffmpeg -i HoneyBee_360p.mp4 -pixel_format yuv420p HoneyBee_360p.yuv % encode INTER with qp 25 ffmpeg -f rawvideo -pix_fmt yuv420p -s:v 352x288 -r 25 -i foreman_cif.yuv -c:v libx264 -qp 25 HoneyBee_360p_qp25.mp4 % INTRA -1- http://ultravideo.cs.tut.fi/#testsequences ffmpeg -f rawvideo -pix_fmt yuv420p -s:v 640x360 -r 25 -i HoneyBee_360p.yuv -c:v libx264 -intra -qp 25 HoneyBee_360p_INTRA_qp25.mp4 {Hint: use FFMPEG software, sudo apt install ffmpeg on linux will install the package for you. User manual is here: https://ffmpeg.org/ffmpeg.html } [1] Obtain the Beauty and Boshphorous sequences in 360p resolution and 4:2:0 YUV format, encode the first 30 frames in INTRA only format, using HEVC codec, and quantization parameter QP=[22, 25, 28 32], or you can use CRF scheme, either is fine. Plot the PSNR of all 30 frames for both sequences (10pts), as well as the SSIM scores (15pts) {hint: CRF based coding: https://trac.ffmpeg.org/wiki/Encode/H.264#crf} [2] Code the 2 sequences with INTER, using HEVC codec, with QP=[22, 25, 28, 32], plot and compare the PSNR and SSIM scores. (25 pts) [3] For the 2nd frame in both sequences, compare the SSIM map for QP = 25 for INTRA vs INTER coding schemes, and explain your observation (50pts) -2- https://ffmpeg.org/ffmpeg.html