Duration: 3:40

3,906 views

171

So this is a follow-up to the previous video (https://www.youtube.com/watch?v=QC4ybDsXBMI ) but this time, the DCT is actually cropped to remove all data outside the upper-left 64x64 pixel region, and then that cropped data is decoded to show the result of the compression.

While JPEG uses DCT as a really important step in image compression, it eliminates data in a different way that is not as simple as "removing data outside the upper-left 64x64 pixel region". So please don't expect this video to look like JPEG. For that, you can simply compress a video using Motion JPEG codec.

By removing data outside the upper-left 64x64 pixel region, we eliminate some fine details from the original image. This slightly blurs the image, and also introduces a compression artifact known as "ringing" where you see faint rings or outlines around sharp edges. Usually, at this compression rate, you would see blockiness start to show on JPEG images, but because JPEG compresses 8x8 pixel blocks of the original image separately and this video compresses the original image as a whole, the blockiness does not show here. The resulting image is blurred, but note that the resolution is exactly the same as the input. The bluriness is introduced by removing data from the discrete-cosine transformed data.

I hope this video was interesting to watch.