Encode a music into an image without notable changes, and decode it

Tools:

OpenCV, scipy.fft, jpeglib, librosa, Pillow, sounddevice

Process Flow:

1. Encoding
   • convert image to YCbCr
   • read audio and fft
   • keep only positive frequencies and filter high freq components
   • round, convert to int16 and then binary, split in group of 2 digits
   • encode real part to Cb domain and imaginary part to Cr domain
   • reconstruct image
2. Decoding
   • split image to YCbCr
   • get the real and imaginary part from Cb and Cr channels
   • merge results to get the fft results in complex terms
   • get full fft and ifft to get audio

Result:

For more information, please refer to docs/ProjectReport.pdf

Future Works:

I love the songs by Khalil Fong(方大同), i wish to also encode his songs in this list. Additionaly, the interpretation and commentary i like about these songs from bilibili.