Why did I get a larger data size when compressing the region of interest than the original video?

Question

I have a medical video, and I want to send it through the network efficiently, so I used the Region of Interest technology to split the original video into the Region of interest and background.

I split the original video and generate the Region of interest and background sub-videos. Encode each sub-video separately using the h265 encoder using the same options. Finally, send them through the network.

But when I compare the original encoded video size with the encoded Region of interest, I found that it has a data size less than the encoded Region of interest.

I'm not an expert in compression. Can anyone help me understand why I got these results, knowing I used the same encoder with the same options?

Answer 1

If I understood correctly, you are splitting your single source video into two sub-sources, corresponding to background and foreground pixels. I don't know how you fill the pixels that fall outside of your regions or interest (background, foreground). But let's assume that you use 0 or the gray value corresponding to the middle of your dynamic range (e.g. 128 for 8bit pixel). But the problem is that you are expecting that a perfect separation of data before compression, would result in at least the same data size as compressing all data in the same source.

This assumption is incorrect for several reasons. But the most important reason is that at the end of the day, you are encoding two video signals, instead of one. Meaning that you will have two sets of headers (which are never free in terms of bitrate). Moreover, those pixels that you filled with 0 (or gray) are also costing you in terms bitrate. Specially if foreground and background areas have complicated shapes, interleaving each other.

One last thing, you should provide your x265 command lines. I fear that you are using constant bitrate (CBR) setting, which means that whatever the content, the encoder will always try to spend a certain number of bits. In that case, for example with a target bitrate of 5Mbps, your setting would result 5Mbps+5Mbps=10Mbps, instead of one single video of 5Mbps.