Introduction to Verilog Method and Implementation of H.264 / AVC
0 Preface
H.264 As a new generation of video compression criteria, H.264 is jointly developed by the Joint Video Panel established by ITU-T video coding expert group and ISO / IEC MPEG (Sport Image Coding Expert Group). Its excellent compression performance will also play an important role in all aspects of digital television broadcast, video real-time communication, network video streaming media transmission, and multimedia communications.
In the core, H.264 / AVC uses various effective technologies, such as unified VLC symbol coding, 1/4 pixel precidity motion estimation, multi-mode motion estimation, based on 4 × 4-piece integer transformation, layered encoding Grammar, etc. Due to the use of integer transformation, the calculation does not have a floating point, and the accuracy is high. These measures make the H.264 algorithm have high coding efficiency. In terms of quantization, the quantization step is 52, and the main analysis is quantified below. Here, a quantified Verilog implementation is proposed, and the transformed data is used as input, and the amount of code stream is used as an output to achieve the purpose of quantization.
1 quantization function introduction
The sampling pulse signal is discrete, but is still continuous in the amplitude and space, that is, the value there may be more than one, which requires the use of a four-round method to use it. The number of values is changed from unlimited multiplexes. This process of turning signal amplitudes from a discrete amount of a discrete amount is called quantization.
Under the premise that the visual effect is not reduced, the quantization process can reduce image coding length and reduce unnecessary information in visual recovery. H.264 uses scalar quantization techniques to map each image sample to a smaller value. The principle of general scale quantifiers is:
Where: y is input sample point coding; QP is quantified step; the quantization value of FQ is Y; the Round () refines the function (which outputs the integer of the input real number).
2 quantization algorithm introduction
In H.264, the quantization step QStep has 52 values. As shown in Table 1. Where QP is a quantitative parameter and is a quantified step number. When QP takes a minimum value 0, the most fine quantization is meant; when the QP takes the maximum value 51, the most rough quantization is represented. Every time the QP increases by 6, QSTEP increases by 1. It can be flexibly selected in this wider quantization step size when applying. For chromaticity encoding, the same quantization step length is generally used in use. In order to avoid color quantization artificial effects when higher quantization steps, the current H.264 draft H. 264 is about 80% of the brightness QP maximum value. The final H.264 draft rule, the maximum value of the brightness QP is 51; the maximum value of the chromaticity QP is 39.
In H.264, the quantization process is to operate on the results of DCT:
Where: yij is a conversion coefficient in the matrix y; Zij is the quantization coefficient of the output; QSTEP is a quantitative step.
The H.264 quantization process must also complete the "EF" multiplication in DCT transform, which can be expressed as:
Where: Wij is the conversion coefficient in the matrix W; PF is an element in the matrix EF. The position (i, j) of the sample point in the image is shown in Table 2.
Using the quantitative step with the quantization parameters increased by 1 to 6, each increase is 6, and the calculation can be further simplified, namely:
Where: floor () is a composite function (its output is not greater than the maximum integer of the input implement). The formula (3) can be written as:
In this way, the MF can take the individual. Table 3 gives an MF value corresponding to the corresponding QP value of 0 to 5. For the case where the QP value is greater than 5, only QBITS value increases by 6 with the QP value, and the corresponding MF value does not change. In this way, the quantization process is an integer operation, and the use of division can be avoided, ensuring that the data is used to deal with the data in the case where the PSNR performance is deteriorated, as shown in Table 3.
The specific quantization process is:
In: "" "" For the right shift operation, the right movement once completed the integer divided by 2; sIGN () as a symbol function; f is the offset. The role of f is to improve the visual effect of the recovery image, such as 2QBITS / 3 to the intra prediction image block F; 2QBITS / 6 is taken to the inter prediction image block.
3 implementation
In this paper, the quantization of H.264 is implemented in Verilog language; simulation using modelsim; use QuartUSII to synthesize.
According to Verilog programming, Modelsim simulation is shown in Figure 1.
The matrix input is [140, -1, -6, 7, -19, -39, 7, -92, 22, 17, 8, 31, -27, -32, -59, -21], and finally quantified The result is [17, 0, -1, 0, -1, -2, 0, -5, 3, 1, 1, 2, -2, -1, -5, -1]. It can be seen that this is in line with the results given by IAIN E.G.Richardson. The development board used is the third generation of red hurricane, and the FPGA chip is Altra EP2C35F484C8. From a comprehensive report, it can be seen that the consumption is less than 1%, as shown in Figure 2. The integrated RTL map is shown in Figure 3.
4 knot
The quantization algorithm of H.264 was introduced and simulated with Modelsim, and the results were exactly the same. The consumption of resources on the FPGA development board is analyzed. It can be seen that the quantization of H.264 can be implemented using FPGA. Technology area
Mei Gao Si Mei Polarfire FPGA device won the "Today Electronics" and 21C.com "2017
Accelerate new technology and drive the future
DSP experts give you a reason to choose FPGA
AccelerComm and Achronix implementation 5G polarization code with SpeedCore EFPGA integration to support customers 5G
2018 Subingisi is a big attack machine learning market
Our other product:
