Difference H264 and H265 algorithms

Question

I try to understand difference of H264 and H265 video compression algorithms. However, I confused the algorithms of uses in H264 and H265. In addition, I found a few differences between these two algorithms.

These differences are

1. H265 has half the channel width of H264. (decrease bandwith)
2. H264 macroblocks has maximum 16x16 size but H265 macroblocks has maximum 64x64 size.
3. H264 has 9 intra-picture, H265 has 35 intra-picture.

However I obsessed these questions.

1. What is the difference inter-intra prediction basically? ( I know that intra prediction predicts from neighboring image data in the same image but inter prediction predicts from reference one or more image data in the one or more images.) I don't visualize in my mind. I don't understand main idea in inter and intra prediction.
2. How do these algorithms use macroblocks?
3. Are there any differences I can add to the list?
4. Are there any basic level resources where I can understand the theory and operation of algorithms at a basic level?

Answer 1

The high-level picture is that h.265 allows lower bandwidth at the same visual quality (or more visual quality at the same bandwidth). It achieves this by using more advanced techniques that require more cpu power and (unfortunately) with some licensing drawbacks.

I think that the fundamental way video codecs works have not changed much over the years. They try to track movement and to express blocks of pixel(-residual) using quantized dct(-ish) coefficients.

Obviously, when the main feature of a codec is compression, and someone have worked really hard to optimize that allready, improving this aspect by a further factor of two means that you had to do something clever.

Inter coding and intra coding is, as you say, about accessing previously encoded data or not. In a typical video stream, there is a lot of temporal redundance, where (nearly) the same was encoded preciously. Basing the current head rendering on the previous version gives a head start and you avoid spending so many bits. One the other hand, if that previous frame was lost due to bot errors (or never received because the user just flipped channels on her tv), then you don’t have that historic data.