H.264 video compression principle

From the perspective of information theory, an image is a source of information, and the data describing the source is the sum of the amount of information (source entropy) and the amount of information redundancy. There are many types of information redundancy, such as spatial redundancy, temporal redundancy, structural redundancy, knowledge redundancy, visual redundancy, etc. Data compression is essentially to reduce these redundancies. It can be seen that reducing redundancy can reduce the amount of data without reducing the amount of information in the source. Mathematically speaking, an image can be regarded as a multidimensional function. Compressing the amount of data describing this function is essentially to reduce its correlation.

According to the components of image information, H.264 uses intra-frame prediction, inter-frame prediction, motion estimation and motion compensation, integer transform and other methods to improve the compression rate of the image. Among them, intra-frame prediction is that H.264 uses the correlation of adjacent pixels based on the possible same nature of adjacent pixels in the image, adopts a new intra-frame prediction mode, predicts through the pixels on the left and above the current pixel block (encoded and reconstructed pixels), and only encodes the difference between the actual value and the predicted value, so that the information of the pixel block encoded in the frame can be expressed with fewer bits; while inter-frame prediction accurately predicts the next frame through methods such as multi-frame reference and smaller motion prediction area, thereby reducing the amount of transmitted data and reducing the time domain correlation of the image. H.264 transforms the residual results of motion estimation and intra-frame prediction from the time domain to the frequency domain, using an integer transform similar to the 4×4 discrete cosine transform (DCT) instead of the floating-point transform of 8×8 DCT as used in MPEG-2 and MPEG-4. The spatial transform based on integers has the advantages of good effect, fast calculation (only addition and shift operations are required), and no adaptation problems will occur in the inverse transformation process. In combination with the quantization process, it ensures that the calculation results have maximum accuracy and will not overflow in a 16-bit computing system. 4×4 transform blocks or 8×8 can further reduce blocking and ringing effects.

The application of H.264 standard provides technical support for video network transmission, especially transmission under low bandwidth (such as wireless transmission). In today's information explosion, the market requires designers to launch products faster and faster, and in order to maintain product competitiveness, they must maintain sufficient advantages in product prices and product application features, especially in the multimedia market. HiSilicon's SoC chip provides designers with the greatest flexibility to design suitable products on the basis of providing standard algorithms, greatly reducing manufacturers' investment costs and shortening the time for manufacturers to launch new products, preparing to win the market.