CN106454354A - AVS2 parallel encoding processing system and method - Google Patents

Abstract

The invention discloses an AVS2 parallel encoding processing system and method, the system includes an encoding system for dividing a frame image into slices with slices as a basic encoding unit, and then performing parallel encoding processing on a plurality of slices obtained after division. unit. The method includes the step of dividing the frame image into slices by taking the slice as a basic coding unit, and then performing parallel encoding processing on a plurality of slices obtained after the division. By using the AVS2 encoding process of the present invention, the encoding process efficiency of AVS2 can be greatly improved, and the requirement of AVS2 real-time encoding can be met. As an AVS2 parallel encoding processing system and method, the present invention can be widely applied in the field of audio and video encoding.

Description

A kind of AVS2 parallel code processing system and method

technical field

The invention relates to audio and video coding and decoding technology, in particular to an AVS2 parallel coding processing system and method.

Background technique

Explanation of technical terms:

TLS: The abbreviation of Thread Local Storage, Chinese is thread local storage.

AVS2 is a new generation of audio and video coding standards independently developed by my country after AVS, and it is the abbreviation of the "Efficient Multimedia Coding for Information Technology" standard. Its goal is: under the condition that the mainstream technology can achieve, when the subjective quality of the reconstructed video is consistent, the coding efficiency of AVS2 for high-definition or higher resolution video is at least 1 times higher than the best performance of AVS1. Under the mainstream encoding configuration, the efficiency is higher than the latest international standard HEVC/H.265. AVS2 and the latest international coding standard HEVC/H. 265 has the same coding efficiency for conventional video, nearly double the coding efficiency of the international standard H.264/AVC and the first-generation national standard AVS1; and for monitoring and other scene videos, the compression efficiency of AVS2 is H. 264/AVC 4 times. Therefore, as the second-generation audio and video codec standard with independent intellectual property rights formulated by my country, AVS2 directly affects my country's core competitiveness in the international video field, and is related to my country's future strategic deployment in the information field. For my country's information industry The rapid development is of great significance. However, since AVS2 introduces a lot of the latest video coding technologies, the computational complexity of coding is also significantly increased, which brings new challenges to the realization of real-time coding.

According to the AVS2 standard, when an AVS2 encoder encodes a frame of image, it takes the LCU (that is, the largest coding unit) as the basic unit to perform serially, and the specific principle is: when encoding each LCU, firstly, the current LCU Initialize the related variables, and then judge whether the LCU belongs to a new stripe, if it belongs to a new stripe, initialize the related variables of the current stripe, and input the stripe header information into the code stream, and then encode the current LCU ; Finally, the code stream to be transmitted is obtained through entropy coding, including strip header data, residual information, block information, etc.; wherein, when storing the code stream, the code stream is continuously superimposed and stored in a total code stream memory , until all LCUs are encoded, the total code stream is output, and one frame of image encoding is completed. It can be seen that the traditional AVS2 encoding processing technology has disadvantages such as low processing efficiency and inability to meet the requirements of AVS2 real-time encoding.

Contents of the invention

In order to solve the above technical problems, the object of the present invention is to provide an AVS2 parallel encoding processing system with high encoding processing efficiency.

Another object of the present invention is to provide an AVS2 parallel encoding processing method with high encoding processing efficiency.

The technical solution adopted in the present invention is: a kind of AVS2 parallel coding processing system, and this system comprises:

The coding unit is configured to divide the frame image into slices by using the slice as a basic coding unit, and then perform parallel coding processing on the divided slices.

Further, the encoding unit includes:

A division module, configured to divide the frame image into slices using the slice as a basic coding unit, so as to obtain multiple slices of the frame image;

The encoding control processing module is used to put all the strips of the current frame image into the task queue in turn, and then use the thread pool to perform parallel encoding processing on the multiple strips.

Further, the use of thread pool to perform parallel encoding processing on multiple stripes specifically includes:

When each stripe is put into the task queue, an idle worker thread in the thread pool is awakened, so that the awakened worker thread encodes the stripe currently put into the task queue;

When there are unencoded stripes in the task queue and all worker threads are busy, create a new worker thread to encode the unencoded stripes, or make the unencoded stripes in the task queue The strip to be encoded is in a waiting state until the end of the task and returns to the worker thread of the thread pool to encode the unencoded strip;

When the worker thread completes the encoding processing of a stripe, it returns to the thread pool again, and adds 1 to the number of stripes that have currently completed encoding processing, and then judges whether there are unencoded stripes in the task queue. If so, Then make the worker thread encode the unencoded strip, otherwise, put the worker thread into a blocking state and wait for the arrival of the next task;

When the number of strips that have completed encoding processing is the same as the total number of strips of the frame image, the main thread is awakened to perform data serial processing.

Further, the encoding processing of the strips is specifically:

The slice header information of the slice is stored, and then the LCUs in the slice are encoded sequentially until all the LCUs in the slice are encoded.

Further, the use of thread pool to perform parallel encoding processing on multiple stripes specifically includes:

When the worker thread needs to access the global variable, it operates on the copy of the global variable stored by itself, so as to realize the access of the global variable.

Further, it also includes a code stream buffer unit, and the code stream buffer unit includes a total code stream memory and a plurality of sub code stream memories;

The total code stream memory is used to store image header information of frame images;

The sub-stream memory is used to store slice header information of a slice and coding information of all LCUs in the slice.

Another technical solution adopted in the present invention is: a kind of AVS2 parallel coding processing method, and this method comprises:

A slice is used as a basic coding unit to divide a frame image into slices, and then parallel encoding processing is performed on a plurality of slices obtained after division.

Further, the step of dividing the frame image into slices by using the slice as the basic coding unit, and then performing parallel encoding processing on the divided slices specifically includes:

dividing the frame image into slices by using the slice as the basic coding unit, so as to obtain multiple slices of the frame image;

Put all the strips of the current frame image into the task queue in turn, and then use the thread pool to encode multiple strips in parallel.

Further, the step of using the thread pool to perform parallel encoding processing on multiple strips specifically includes:

When each stripe is put into the task queue, an idle worker thread in the thread pool is awakened, so that the awakened worker thread encodes the stripe currently put into the task queue;

When there are unencoded stripes in the task queue and all worker threads are busy, create a new worker thread to encode the unencoded stripes, or make the unencoded stripes in the task queue The strip to be encoded is in a waiting state until the end of the task and returns to the worker thread of the thread pool to encode the unencoded strip;

When the worker thread completes the encoding processing of a stripe, it returns to the thread pool again, and adds 1 to the number of stripes that have currently completed encoding processing, and then judges whether there are unencoded stripes in the task queue. If so, Then make the worker thread encode the unencoded strip, otherwise, put the worker thread into a blocking state and wait for the arrival of the next task;

When the number of strips that have completed encoding processing is the same as the total number of strips of the frame image, the main thread is awakened to perform data serial processing.

Further, the step of using the thread pool to perform parallel encoding processing on multiple strips specifically includes:

When the worker thread needs to access the global variable, it operates on the copy of the global variable stored by itself, so as to realize the access of the global variable.

The beneficial effect of the present invention is that: the system of the present invention divides the frame image into slices using slices as the basic coding unit, and then performs parallel coding processing on the multiple slices obtained after the division, thereby realizing the coding process of the frame image Therefore, compared with the traditional AVS2 encoding process, the system of the present invention does not need to judge each LCU in the encoding process to determine whether the LCU belongs to a new strip, which can save a lot of repeated redundant steps, Greatly improve the efficiency of AVS2 encoding processing. Moreover, the system of the present invention encodes multiple strips in a data parallel processing manner. Therefore, compared with the traditional serial method, the data processing efficiency of the system of the present invention is higher, and it can better meet the requirements of AVS2 real-time encoding. needs.

Another beneficial effect of the present invention is that: the method of the present invention divides the frame image into slices using the slice as the basic coding unit, and then performs parallel encoding processing on the multiple slices obtained after division, thereby realizing the encoding of the frame image Therefore, compared with the traditional AVS2 encoding process, the method of the present invention does not need to judge each LCU in the encoding process to determine whether the LCU belongs to a new strip, which can save a lot of redundant redundancy steps, and the method of the present invention adopts parallel mode to carry out strip coding, so by using the method of the present invention, the coding processing efficiency of AVS2 can be greatly improved to meet the needs of AVS2 real-time coding.

Description of drawings

The specific embodiment of the present invention will be further described below in conjunction with accompanying drawing:

Fig. 1 is a structural block diagram of a kind of AVS2 parallel coding processing system of the present invention;

FIG. 2 is a flow chart of the steps of an AVS2 parallel encoding processing method of the present invention.

detailed description

As shown in Figure 1, an AVS2 parallel encoding processing system, the system includes:

The coding unit is configured to divide the frame image into slices by using the slice as a basic coding unit, and then perform parallel coding processing on the divided slices. The encoding unit can preferably be applied in an AVS2 encoder.

As a preferred implementation of the system in this embodiment, the encoding unit includes:

A division module, configured to divide the frame image into slices using the slice as a basic coding unit, so as to obtain multiple slices of the frame image;

The encoding control processing module is used to put all the strips of the current frame image into the task queue in turn, and then use the thread pool to perform parallel encoding processing on the multiple strips.

As a preferred implementation of the system of this embodiment, the parallel encoding processing of multiple stripes by using the thread pool specifically includes:

When each stripe is put into the task queue, an idle worker thread in the thread pool is awakened, so that the awakened worker thread encodes the stripe currently put into the task queue;

When there are unencoded stripes in the task queue and all worker threads are busy, create a new worker thread to encode the unencoded stripes, or make the unencoded stripes in the task queue The strip to be encoded is in a waiting state until the end of the task and returns to the worker thread of the thread pool to encode the unencoded strip;

When the worker thread completes the encoding processing of a stripe, it returns to the thread pool again, and adds 1 to the number of stripes that have currently completed encoding processing, and then judges whether there are unencoded stripes in the task queue. If so, Then make the worker thread encode the unencoded strip, otherwise, put the worker thread into a blocking state and wait for the arrival of the next task;

When the number of strips that have completed encoding processing is the same as the total number of strips of the frame image, the main thread is awakened to perform data serial processing.

As a preferred implementation of the system in this embodiment, the encoding processing of the strips is specifically:

The slice header information of the slice is stored, and then the LCUs in the slice are encoded sequentially until all the LCUs in the slice are encoded.

As a preferred implementation of the system of this embodiment, the use of the thread pool to perform parallel encoding processing on multiple stripes specifically includes:

When the worker thread needs to access the global variable, it operates on the copy of the global variable stored by itself, so as to realize the access of the global variable.

As a preferred implementation of the system of this embodiment, the use of the thread pool to perform parallel encoding processing on multiple stripes specifically includes:

Create a thread pool, and create a plurality of worker threads in the thread pool, usually, the number of the worker threads defaults to the number of cores of the current processor; then, put the plurality of worker threads into the idle thread queue.

Preferably, for the worker thread, it is placed in a blocked state when it does not need to perform tasks, and it is placed in a schedulable state when it needs to perform tasks, so that it will not occupy the processing when the worker thread has no work items. Server resources, only occupying a small amount of memory space, greatly improving the flexibility of operation.

As a preferred implementation of the system of this embodiment, it also includes a code stream buffer unit, and the code stream buffer unit includes a total code stream memory and a plurality of sub code stream memories;

The total code stream memory is used to store image header information of frame images;

The sub-stream memory is used to store slice header information of a slice and coding information of all LCUs in the slice. After all the strips are encoded, the code streams stored in each sub-stream memory are sequentially combined into the total code stream according to the order of the stripes, and the finally obtained code stream is a coded frame image code stream.

Preferably, the number of sub-stream memories is N, and the allocated memory size of each sub-stream memory is 1/N of the memory size of the total code stream memory.

As shown in Figure 2, an AVS2 parallel encoding processing method includes:

A slice is used as a basic coding unit to divide a frame image into slices, and then parallel encoding processing is performed on a plurality of slices obtained after division.

As a preferred implementation of the method in this embodiment, the step of dividing the frame image into slices using slices as the basic coding unit, and then performing parallel encoding processing on the divided slices specifically includes:

dividing the frame image into slices by using the slice as the basic coding unit, so as to obtain multiple slices of the frame image;

Put all the strips of the current frame image into the task queue in turn, and then use the thread pool to encode multiple strips in parallel.

As a preferred implementation of the method in this embodiment, the step of using a thread pool to perform parallel encoding processing on multiple stripes specifically includes:

When each stripe is put into the task queue, an idle worker thread in the thread pool is awakened, so that the awakened worker thread encodes the stripe currently put into the task queue;

When there are unencoded stripes in the task queue and all worker threads are busy, create a new worker thread to encode the unencoded stripes, or make the unencoded stripes in the task queue The strip to be encoded is in a waiting state until the end of the task and returns to the worker thread of the thread pool to encode the unencoded strip;

When the worker thread completes the encoding processing of a stripe, it returns to the thread pool again, and adds 1 to the number of stripes that have currently completed encoding processing, and then judges whether there are unencoded stripes in the task queue. If so, Then make the worker thread encode the unencoded strip, otherwise, put the worker thread into a blocking state and wait for the arrival of the next task;

When the number of strips that have completed encoding processing is the same as the total number of strips of the frame image, the main thread is awakened to perform data serial processing.

As a preferred implementation of the method in this embodiment, the encoding processing of the strips is specifically:

The slice header information of the slice is stored, and then the LCUs in the slice are encoded sequentially until all the LCUs in the slice are encoded.

As a preferred implementation of the method in this embodiment, the step of using the thread pool to perform parallel encoding processing on multiple stripes specifically includes:

When the worker thread needs to access the global variable, it operates on the copy of the global variable stored by itself, so as to realize the access of the global variable.

As a preferred implementation of the method in this embodiment, the step of using the thread pool to perform parallel encoding processing on multiple stripes specifically includes:

Create a thread pool, and create a plurality of worker threads in the thread pool, usually, the number of the worker threads defaults to the number of cores of the current processor; then, put the plurality of worker threads into the idle thread queue.

A specific embodiment of the invention

In the AVS2 standard, when encoding a frame of image, each slice will initialize its binary symbol model and entropy encoder before starting encoding, and each LCU in the slice will only refer to this slice when encoding Other LCUs will not use the data of other strips in the frame image. Therefore, it can be seen that the encoding between the slices in the frame image is independent of each other, so that the parallel encoding process can be realized by using the slice as the basic encoding unit, which can achieve the purpose of rapidly increasing the encoding speed.

The AVS2 video parallel coding optimization algorithm based on slices as the basic coding unit specifically includes the following four components:

1. Architecture Design of Parallel Encoder

If you want to implement slice-based AVS2 parallel encoding, the encoder framework must first use slices as the basic coding unit for encoding, which requires adjustments to the encoding framework in the original encoder with LCU as the basic unit. The adjusted encoder, when encoding a frame of image, uses the strip as the basic coding unit for encoding. After all the LCUs in a strip are encoded, the next strip is encoded until the image in the frame All strip encoding is done. When encoding a slice, first initialize the slice and store the slice header information of the slice, so that it is no longer necessary to determine whether each LCU belongs to a new slice when encoding each LCU as before. After the slice is initialized, the LCUs in the slice are encoded sequentially until all the LCUs in the slice are encoded.

It can be seen from the above that the coding system in this embodiment may be specifically an AVS2 coder, and the coder includes a coding unit, and the coding unit is specifically used to divide a frame image into slices using slices as a basic coding unit, Then perform parallel encoding processing on the multiple strips obtained after division;

Wherein, the encoding processing of the slice specifically includes: initializing the slice and storing header information of the slice, and then encoding the LCUs in the slice in sequence until the encoding of all the LCUs in the slice is completed.

2. Stream buffer design

On the basis of the adjusted parallel encoding framework, the code stream buffer needs to be adjusted. When encoding a frame of images containing N strips, the encoder establishes a total code stream memory (GlobalBitStream) and N sub-bitstream memories (SliceBitStream[i]), and allocate a certain memory space for each memory, wherein the memory size allocated to the sub-bitstream memory is 1/N of the memory size allocated to the total bitstream memory. In the encoding process, the image header information is first stored in the total code stream memory, and then when encoding each slice, the slice header information of each slice and the encoding information of all LCUs in the slice are stored in the code stream buffer The slice header information of a slice and the encoding information of all LCUs in the slice are stored in the corresponding sub-stream memory, which is equivalent to a sub-stream storage. After all the strips are encoded, the code streams in each sub-stream memory are merged into the total code stream according to the order of the stripes, and the finally obtained code stream is the code stream for encoding one frame of image.

It can be seen from the above that the encoder in this embodiment further includes a code stream buffer unit, and the code stream buffer unit includes a total code stream memory and a plurality of sub code stream memories;

The total code stream memory is used to store image header information of frame images;

The sub-stream memory is used to store slice header information of a slice and coding information of all LCUs in the slice.

Wherein, the memory size allocated to the sub-stream memory is 1/N of the memory size allocated to the total code stream memory, and N is the total number of sub-stream memories.

3. Thread pool design

The thread pool in this embodiment is a form of multi-thread processing, which adopts the technique of creating threads in advance. By using the thread pool technology to encode multiple strips in parallel, it can save the CPU time that needs to be continuously created and destroyed when encoding each strip. Specifically, the specific steps of using the thread pool to perform parallel encoding processing on multiple strips include:

1. After the AVS2 encoder starts, first create a thread pool, and create a certain number of worker threads in the thread pool (the default is the number of cores of the current processor), and then put these worker threads into an idle thread queue; and When there are no work items, these worker threads are blocked, so that when there are no work items, the worker threads do not occupy processor resources, only occupy a small amount of memory space, and are schedulable when there are work items;

2. When the main program executes to encode a frame of image, first put all the slices of the current frame of image into a task queue, QueueWork (encode_one_slice, (void *) & CurrentMbNumber), when all the slices are put After entering the task queue, create an event object, whose event signal is set to no signal state, WorkFinished = CreateEvent (NULLTRUE, FALSE, NULL); then, the main thread is set to blocked state, WaitForSingleObject (WorkFinished, INFINITE), waiting for the event The signal changes to the signaled state;

3. At the same time, when each stripe is placed in the task queue, an idle worker thread in the thread pool is awakened, and the worker thread starts encoding the current stripe, and the function is used when calling the task in the worker thread Pointers and function pointers are called as parameters to pass relevant data, and the task function used for calling is defined as void ThreadFun (void *context); when there are unencoded strips in the task queue, and all worker threads are busy State, that is, when there are no idle threads to perform encoding processing, the thread pool manager can create a certain number of additional worker threads to perform more tasks, or let the stripes waiting for encoding in the task queue be in the waiting state first, waiting After the task is finished, return to the thread of the thread pool for execution;

4. Each worker thread completes the encoding of a stripe, then returns to the thread pool, and adds 1 to the number of completed stripes; while the worker thread returning to the thread pool does not need to exit, if the task queue still exists unexecuted Encoded strips, the worker thread continues to encode the strips, otherwise it is set to a blocked state and waits for the arrival of the next task; when the number of encoded strips is the same as the total number of strips in the current frame image , then the event object, its event signal WorkFinished is set to a signal state, SetEvent (WorkFinished), thereby waking up the main thread, and the main thread enters the next serial part, including synchronous data and merging streams, etc.;

5. After the encoding of one frame of image is completed, if the video sequence is still not over, return to step 2 to continue encoding. If the video sequence ends, the thread is destroyed and the thread pool is destroyed.

It can be seen from the above that the encoding control processing module included in the encoding unit in this embodiment is specifically used to put all the slices of the current frame image into the task queue in sequence, and then use the thread pool to perform parallel encoding processing on multiple slices.

Wherein, the parallel encoding processing of multiple strips by using the thread pool specifically includes:

Create a thread pool, and create a plurality of working threads in the thread pool, usually, the number of the working threads can preferably default to the number of cores of the current processor; then, put the multiple working threads into the idle thread queue Among them, when the worker thread does not need to perform tasks, it is set to a blocked state, and when it needs to perform tasks, it is set to a schedulable state;

When performing frame image encoding, put all the strips in the current frame image into a task queue in turn; and when all the strips are put into the task queue, create an event object and set its event signal to No signal state, then set the main thread to a blocked state, and wait for the event signal to turn into a signal state;

When each stripe is put into the task queue, an idle worker thread in the thread pool is awakened, so that the awakened worker thread encodes the stripe currently put into the task queue;

When there are unencoded stripes in the task queue and all worker threads are busy, create a new worker thread to encode the unencoded stripes, or make the unencoded stripes in the task queue The strip to be encoded is in a waiting state until the end of the task and returns to the worker thread of the thread pool to encode the unencoded strip;

When the worker thread completes the encoding processing of a stripe, it returns to the thread pool again, and adds 1 to the number of stripes that have currently completed encoding processing, and then judges whether there are unencoded stripes in the task queue. If so, Then make the worker thread encode the unencoded strip, otherwise, put the worker thread into a blocking state and wait for the arrival of the next task;

When the number of stripes that have been encoded is the same as the total number of stripes in the frame image, wake up the main thread for subsequent data serial processing, such as subsequent data synchronization processing, code stream merging processing, etc.;

After finishing the encoding processing of a frame image, then judge whether the frame images contained in the video sequence have been encoded, if so, it means that the encoding processing of the video sequence has ended, at this time, then destroy the thread and destroy the thread pool; otherwise , then put all strips of a frame of image that have not been encoded into the task queue in turn, so that the thread pool is used to encode multiple strips.

4. Implementation of local data storage

In this embodiment, the AVS2 encoder utilizes the TLS technology to realize simultaneous access to shared resources by different working threads. After using TLS, the worker threads of parallel encoding can access global variables at the same time, and can enjoy global variables independently, that is to say, each worker thread using TLS can have a copy of global variables, and only To operate on this copy, the threads are independent of each other and do not interfere with each other, so this can further improve the work efficiency of parallel encoding processing. The implicit use of TLS is realized by the cooperation of the compiler, loader and linker. The specific operation is to add the modifier __declspec(thread) before the variables that need to use thread local storage. This method is relatively simple, and more suitable for a large number of local storage variables and frequent calls in the AVS2 encoder.

Preferably, after a deep understanding of the key modules in AVS2 encoding and detailed analysis of the specific codes of the encoder, this algorithm sets thread local storage for a total of 155 global variables, first creates the TlsData data structure, and stores the 155 required Put the thread-locally stored variables into it, and then execute the __declspec(thread) TlsData *TlsInfo operation to generate a thread-locally stored pointer. Every time you need to use a certain variable, such as current_mb_nr, you only need to execute TlsInfo-> current_mb_nr means that the variable used by the current working thread is current_mb_nr, so that there will be no influence between working threads. In addition, after the main thread puts the encoding tasks of each strip into the task queue, the worker thread receives the encoding tasks, first allocates a memory space of TlsData size to the *TlsInfo pointer, and then uses the tlsmalloc() function to locally store the encoding tasks for all threads. Variables allocate the required memory space, and finally use the tlsfree() function to release the memory space allocated in all worker threads after encoding a strip.

As can be seen from the above, the use of the thread pool to perform parallel encoding processing on multiple stripes specifically includes:

When the worker thread needs to access the global variable, it operates on the copy of the global variable stored by itself, so as to realize the access of the global variable;

And this step specifically includes:

Set up thread-local storage for corresponding multiple global variables. Specifically, first create a TlsData data structure, put multiple global variables into it, and then execute the __declspec(thread) TlsData *TlsInfo operation to generate a thread-local storage A pointer; wherein, the global variable is preceded by a modifier __declspec(thread);

When the worker thread receives the stripe encoding task, it first allocates a memory space of TlsData size to the *TlsInfo pointer, and then uses the tlsmalloc() function to allocate the required memory space for all thread-locally stored variables;

When the worker thread accesses the global variable, execute the instruction TlsInfo->variable;

After the worker thread has finished encoding the stripe, use the tlsfree() function to release the memory space allocated in all worker threads.

The above is a specific description of the preferred implementation of the present invention, but the invention is not limited to the described embodiments, and those skilled in the art can also make various equivalent deformations or replacements without violating the spirit of the present invention. , these equivalent modifications or replacements are all within the scope defined by the claims of the present application.

Claims (10)

1. A kind of AVS2 parallel coding processing system, it is characterized in that: the system comprises: The coding unit is configured to divide the frame image into slices by using the slice as a basic coding unit, and then perform parallel coding processing on the divided slices. 2. a kind of AVS2 parallel encoding processing system according to claim 1, is characterized in that: described encoding unit comprises: A division module, configured to divide the frame image into slices using the slice as a basic coding unit, so as to obtain multiple slices of the frame image; The encoding control processing module is used to put all the strips of the current frame image into the task queue in turn, and then use the thread pool to perform parallel encoding processing on the multiple strips. 3. a kind of AVS2 parallel coding processing system according to claim 2, is characterized in that: described utilize thread pool to carry out parallel coding processing to a plurality of stripes, it specifically comprises: When each stripe is put into the task queue, an idle worker thread in the thread pool is awakened, so that the awakened worker thread encodes the stripe currently put into the task queue; When there are unencoded stripes in the task queue and all worker threads are busy, create a new worker thread to encode the unencoded stripes, or make the unencoded stripes in the task queue The strip to be encoded is in a waiting state until the end of the task and returns to the worker thread of the thread pool to encode the unencoded strip; When the worker thread completes the encoding processing of a stripe, it returns to the thread pool again, and adds 1 to the number of stripes that have currently completed encoding processing, and then judges whether there are unencoded stripes in the task queue. If so, Then make the worker thread encode the unencoded strip, otherwise, put the worker thread into a blocking state and wait for the arrival of the next task; When the number of strips that have completed encoding processing is the same as the total number of strips of the frame image, the main thread is awakened to perform data serial processing. 4. a kind of AVS2 parallel coding processing system according to claim 3, is characterized in that: described strip is carried out coding processing, and it is specifically: The slice header information of the slice is stored, and then the LCUs in the slice are encoded sequentially until all the LCUs in the slice are encoded. 5. a kind of AVS2 parallel coding processing system according to claim 3, is characterized in that: described utilize thread pool to carry out parallel coding processing to a plurality of stripes, it specifically also comprises: When the worker thread needs to access the global variable, it operates on the copy of the global variable stored by itself, so as to realize the access of the global variable. 6. A kind of AVS2 parallel encoding processing system according to any one of claims 1-5, characterized in that: it also includes a code stream buffer unit, and the code stream buffer unit includes a total code stream memory and a plurality of sub code stream memory ; The total code stream memory is used to store image header information of frame images; The sub-stream memory is used to store slice header information of a slice and coding information of all LCUs in the slice. 7. An AVS2 parallel encoding processing method, characterized in that: the method comprises: A slice is used as a basic coding unit to divide a frame image into slices, and then parallel encoding processing is performed on a plurality of slices obtained after division. 8. A kind of AVS2 parallel coding processing method according to claim 7, it is characterized in that: described frame image is carried out strip division with strip as basic coding unit, then a plurality of strips obtained after division are carried out parallel coding This step specifically includes: dividing the frame image into slices by using the slice as the basic coding unit, so as to obtain multiple slices of the frame image; Put all the strips of the current frame image into the task queue in turn, and then use the thread pool to encode multiple strips in parallel. 9. A kind of AVS2 parallel coding processing method according to claim 8, it is characterized in that: described utilize thread pool to carry out this step of parallel coding processing to a plurality of stripes, it specifically comprises: When each stripe is placed in the task queue, an idle worker thread in the thread pool is awakened, and the awakened worker thread encodes the stripe currently put into the task queue; When there are unencoded stripes in the task queue and all worker threads are busy, create a new worker thread to encode the unencoded stripes, or make the unencoded stripes in the task queue The strip to be encoded is in a waiting state until the end of the task and returns to the worker thread of the thread pool to encode the unencoded strip; When the worker thread completes the encoding processing of a stripe, it returns to the thread pool again, and adds 1 to the number of stripes that have currently completed encoding processing, and then judges whether there are unencoded stripes in the task queue. If so, Then make the worker thread encode the unencoded strip, otherwise, put the worker thread into a blocking state and wait for the arrival of the next task; When the number of strips that have completed encoding processing is the same as the total number of strips of the frame image, the main thread is awakened to perform data serial processing. 10. A kind of AVS2 parallel coding processing method according to claim 9, it is characterized in that: described utilize thread pool to carry out this step of parallel coding processing to a plurality of stripes, it specifically also comprises: When the worker thread needs to access the global variable, it operates on the copy of the global variable stored by itself, so as to realize the access of the global variable.