WO2022205297A1 - Data processing method and device, chip, unmanned aerial vehicle, and storage medium - Google Patents

Abstract

A data processing method and device, a chip, an unmanned aerial vehicle, and a storage medium. The data processing method comprises: obtaining an image to be processed of a preset data format, and storing data of said image according to a preset storage format (101); obtaining a preset number of first input data from the data according to the preset data format, the preset number being less than the number of input data satisfying a preset transformation algorithm (102); on the basis of the obtained first input data, directly multiplexing, from back to front, a preset number of second input data in the previous group of input data satisfying the preset transformation algorithm, and using the second input data and the first input data as input data to be processed (103); and performing calculation according to the input data to be processed and the preset transformation algorithm to obtain a processing result indicating data frequency of said image (104). The input data to be processed is obtained on the basis of the first input data and the directly multiplexed second input data, and extra cache is not needed, such that the efficiency of a calculation process can be improved.

Description

Data processing method, device, chip, drone and storage medium technical field

The present invention relates to the field of image data processing, and in particular, to a data processing method, device, chip, drone and storage medium.

Background technique

With the continuous development of imaging technology, system-on-chip (SOC) bandwidth is more and more likely to become the bottleneck of data transmission. Therefore, image compression technology is widely used. DWT algorithm, DWT (Discrete Wavelet Transformation), is usually used in the process of image compression and encoding, namely discrete wavelet transform. As the front-end part of the Huming compression algorithm, the performance of the DWT algorithm also has an important impact on the performance of the entire compression algorithm, and how to improve the performance of the algorithm becomes very important.

SUMMARY OF THE INVENTION

The present invention provides a data processing method, equipment, chip, unmanned aerial vehicle and storage medium, which are used to improve the efficiency of the DWT algorithm.

A first aspect of the present invention is to provide a data processing method, comprising: acquiring a to-be-processed image in a preset data format, and storing the data of the to-be-processed image according to a preset storage format; and according to the preset data format, from Obtain a preset number of first input data from the data, and the preset number is less than the number of input data that conforms to the preset transformation algorithm; based on the obtained first input data, according to the order from back to front Directly multiplex the last set of second input data that conforms to the preset number in the input data of the preset transformation algorithm, and use the first input data as input data to be processed; according to the input data to be processed and the preset input data The transformation algorithm is used for calculation, and the processing result representing the data frequency of the image to be processed is obtained.

A second aspect of the present invention is to provide a data processing device, comprising: a memory and a processor; the memory is used to store a computer program; the processor invokes the computer program to implement the following steps: obtaining a preset For the image to be processed in the data format, the data of the to-be-processed image is stored according to the preset storage format; according to the preset data format, a preset number of first input data is obtained from the data, and the preset number is small In accordance with the number of input data in accordance with the preset transformation algorithm; Based on the obtained first input data, directly multiplex the previous group of data in accordance with the preset number in the preset transformation algorithm input data in the order from back to front. The second input data and the first input data are used as input data to be processed; calculation is performed according to the input data to be processed and the preset transformation algorithm to obtain a processing result representing the data frequency of the image to be processed.

A third aspect of the present invention is to provide a chip, comprising: the device described in the second aspect.

A fourth aspect of the present invention is to provide an unmanned aerial vehicle, comprising: an airframe and the device according to the second aspect.

A fifth aspect of the present invention is to provide a computer-readable storage medium, the storage medium is a computer-readable storage medium, and program instructions are stored in the computer-readable storage medium, and the program instructions are used in the first aspect. method described.

According to the present invention, the image to be processed in the preset data format is acquired, and the data of the image to be processed is stored according to the preset storage format; and the preset number of first input data is acquired from the data according to the preset data format, and the preset number is less than The number of input data that conforms to the preset transformation algorithm; based on the obtained first input data, the second input data of the preset number of the previous group of input data that conform to the preset transformation algorithm are directly multiplexed in the order from back to front, and take the first input data as input data to be processed; perform calculation according to the input data to be processed and a preset transformation algorithm to obtain a processing result representing the data frequency of the image to be processed.

Wherein, based on the obtained first input data, the previous set of second input data that conforms to the preset number in the input data of the preset transformation algorithm are directly multiplexed in the order from back to front, and are treated with the first input data as pending processing. Enter data and perform subsequent data calculations. Since the input data to be processed is obtained based on the first input data and the directly multiplexed second input data, no additional buffering is required, thereby improving the efficiency of the calculation process and the performance of corresponding algorithms, such as the DWT algorithm.

Description of drawings

The drawings described herein are used to provide further understanding of the present application and constitute a part of the present application. The schematic embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation of the present application. In the attached image:

1 is a schematic flowchart of a data processing method according to an embodiment of the present invention;

2 is a schematic diagram of image processing provided by an embodiment of the present invention;

3 is a schematic diagram of a data format of an image YUV420 provided by an embodiment of the present invention;

Fig. 4 is the schematic diagram of the ping-pong buffer storage data provided by the embodiment of the present invention;

5 is a schematic diagram of data storage after preprocessing provided by an embodiment of the present invention;

6 is a schematic diagram of a specific storage result of data provided by an embodiment of the present invention;

7 is a schematic diagram of data calculation provided by an embodiment of the present invention;

8 is a schematic diagram of a calculation result provided by an embodiment of the present invention;

9 is a schematic diagram of a calculation result provided by an embodiment of the present invention;

10 is a schematic structural diagram of a data processing apparatus according to an embodiment of the present invention;

FIG. 11 is a schematic structural diagram of a data processing device according to an embodiment of the present invention.

Detailed ways

In order to make the purposes, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments These are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terms used herein in the description of the present invention are for the purpose of describing specific embodiments only, and are not intended to limit the present invention.

In order to facilitate the understanding of the technical solutions and technical effects of the present application, the prior art is briefly described below:

As can be seen from the foregoing, because the bandwidth of the system on chip (SOC) is more and more likely to become the bottleneck of data transmission, the image compression technology is widely used. Among them, the DWT algorithm needs additional storage in the middle of the calculation process, and then read back to participate in the calculation, which is not efficient.

Therefore, the embodiments of the present application provide a data processing method, which can improve the performance of the algorithm.

Some embodiments of the present invention will be described in detail below with reference to the accompanying drawings. The following embodiments and features in the embodiments may be combined with each other without conflict between the embodiments.

1 is a schematic flowchart of a data processing method provided by an embodiment of the present invention; the method 100 provided by an embodiment of the present application may be executed by a flying device, such as an unmanned aerial vehicle, and more specifically, a system-on-chip SOC, such as this System-on-chip encoders, etc. The method 100 includes the following steps:

101: Acquire a to-be-processed image in a preset data format, and store data of the to-be-processed image according to a preset storage format.

102: Acquire a preset amount of first input data from the data according to the preset data format.

Wherein, the preset number is less than the number of input data conforming to the preset transformation algorithm.

103: Based on the obtained first input data, directly multiplex the last set of second input data that conforms to the preset number in the input data of the preset transformation algorithm in the order from back to front, and use the first input data as pending processing Input data.

104 : Perform calculation according to the input data to be processed and a preset transformation algorithm to obtain a processing result representing the data frequency of the image to be processed.

It should be noted that the execution body may also be other intelligent terminals, such as mobile phones, computers, servers, and the like.

The above steps are described in detail below:

101: Acquire a to-be-processed image in a preset data format, and store data of the to-be-processed image according to a preset storage format.

Wherein, the preset data format may refer to the data format of the preset image record, such as the YUV preset data format, and more specifically may be YUV420. YUV, divided into three components, "Y" represents the brightness (Luminance or Luma), that is, the gray value; while "U" and "V" represent the chroma (Chrominance or Chroma), which is used to describe the image. Hue and Saturation, used to specify the color of the pixel.

The preset storage format may refer to a preset storage format set based on the preset data format. For example, it can be stored in the corresponding buffer space according to the order of each component of YUV.

For example, when a drone is sailing in the sky, images can be taken through the camera mounted on the drone. As shown in Figure 2, the image signal processor ISP201 in the system-on-chip SOC is responsible for receiving the original signal data of the sensor (Sensor) of the camera, and then processing to obtain a visible image. The image is then sent to an FBC (Frame Buffer Compression, frame buffer compression) encoder, namely the FBC encoder 202, and is calculated by a subset module DWT algorithm on the encoder to compress the image. The compressed image stream is input into the DDR (Double Data Rate Synchronous Dynamic Random Access Memory, double data rate synchronous dynamic random access memory) memory 203, thereby reducing the write bandwidth. Then the corresponding FBC decoder, that is, the FBC decoder 204 requests the compressed image stream from the bus, decodes it, and outputs the decoded image data to the other modules 205 at the next level, thereby reducing the read stream. Other modules 205 can process the decompressed image data, such as a display module, etc., to display the image.

In the above encoder, the image to be processed: YUV420 image data can be received through the AXI protocol (Advanced eXtensible Interface, a bus protocol). As shown in Figure 3, Figure 3 shows the data format of the YUV420 image data. It is then cached to the internal random access memory ram. The cache process is written according to the bit width of 128 bits (the AXI bus is 128 bits wide). For an 8bit image (as shown in Figure 3), the image data in the internal ram is tightly arranged because 128bit contains 16 data. Then, according to the sequence of each component of YUV in the YUV420 image data, it can be stored in a buffer area buffer.

It should be noted that the UAV can also receive the original image sent by other devices, so it can be encoded by the above encoder directly through the AXI protocol without going through the image signal processor ISP201. It should be understood that the image transmission manner shown in FIG. 2 can be adjusted according to different application scenarios. For example, other modules may also be MCTF (Motion Compensated Temporal-Filtering, motion compensation temporal filtering) modules, which can read the encoded code stream directly from the DDR memory 203 . Other forms will not be repeated too much.

In order to improve the image compression process, especially the performance of the DWT algorithm, two buffer cache areas of symmetric storage structures can be set, such as pingpong buffer (a means of data caching, which defines two cache areas of symmetric storage structures, one is ping buffer, the other is pong buffer). This cache area is implemented by register files.

Specifically, obtaining the image to be processed in the preset data format, and storing the data of the image to be processed according to the preset storage format includes: acquiring the image in the YUV preset data format as the image to be processed; The group of YUV component data is sequentially buffered in the corresponding buffer space in the first buffer area; after the first group of YUV component data is buffered, the second group of YUV component data is sequentially buffered in the corresponding buffer space in the second buffer area.

Wherein, the first cache area may be a ping buffer, and the second cache area may be a pong buffer. It should be understood that two other symmetric storage structure buffer cache areas may also be set to implement double buffer cache storage.

For example, according to the foregoing, after the encoder obtains the YUV420 image data, as shown in Figure 3, the line Y0 is stored in Y0403 in the ping buffer401 shown in Figure 4. Store the row Y1 in Figure 3 into Y1 in the ping buffer401 shown in Figure 4. Store the UV line after the Y1 line in Figure 3 into the UV in the ping buffer401 shown in Figure 4. Thus, the ping buffer401 is stored. Then store the pong buffer 402 as shown in Figure 4. At this time, it can be stored from the Y2 row in Figure 3, and so on, and will not be repeated. The ping-pong buffer is stored.

Since two buffer cache areas of a symmetric storage structure are used, subsequent data processing steps, such as the DWT algorithm, can be started after the storage in one of the buffer cache areas is completed. After the processing of the data in the current buffer cache area is completed, since another buffer cache area has also completed storage, the calculation can be performed directly on the other buffer cache area without waiting for data, thereby relatively improving the calculation speed.

Specifically, the method 100 further includes: after buffering the first group of YUV component data, performing a step of obtaining a preset number of first input data from the data according to a preset data format to obtain a processing result, and continuing to cache the corresponding data The YUV component data of the group is sent to the first buffer area, until the YUV component data is buffered; after the second group of YUV component data is buffered, the step of obtaining a preset number of first input data from the data is performed according to the preset data format. , so as to obtain the processing result, continue to buffer the YUV component data of the corresponding group in the second buffer area until the YUV component data is buffered.

It can be known from the foregoing that, for example, after storing the ping buffer 401 shown in FIG. 4 , the DWT subset module in the encoder can start data processing, that is, data calculation, and obtain the corresponding calculation result. Then calculate the data in the pong buffer 402 as shown in Figure 4, and obtain the corresponding calculation result. At the same time, continue to store the ping buffer401. After processing the data in pong buffer402, calculate the data in the new ping buffer401. And so on until the processing of the image data is completed.

In addition, it can be known from the foregoing that the original image data is data of a preset bit depth type, such as 8-bit image data. In order to be compatible with more different data, such as 8-bit image data or 10-bit image data, etc., the original image data may be preprocessed (as described in the following embodiments) to satisfy different types of image data. In this way, the buffer space can buffer data of images to be processed with at least two different bit depths.

Specifically, the method 100 further includes: based on the data of the image to be processed with the largest bit depth among at least two different bit depths, expanding the corresponding original preset cache space in any cache area to generate a corresponding cache space.

Wherein, the original preset buffer space buffers the data of the image to be processed with the smallest bit depth among at least two different bit depths.

For example, according to the foregoing, for 8-bit or 10-bit image data in 10-bit, there is splicing at the boundary of the internal ram. For example, the data volume of 10bitY0 or Y1 is 64*1*10bit=640bit respectively. Considering the convenience of subsequent data entry, a 160-bit storage bit width is proposed here. For 8-bit image data, each time 128-bit data is received from the AXI bus, it is preprocessed and expanded to 160-bit.

It should be understood that, for the data of the images to be processed with at least two different bit depths, the storage bit width can be expanded in the above manner, so as to accommodate the data of the images to be processed with different bit depths.

For the data of the image to be processed with a non-maximum bit depth, data expansion can be used to satisfy the above-mentioned expansion of the storage bit width.

Specifically, the method 100 further includes: performing data expansion on the data of the to-be-processed image with a non-maximum bit depth among the at least two different bit depths, to obtain the data of the to-be-processed image with the maximum bit depth, so as to be cached in the buffer space.

For example, according to the foregoing, for the above-mentioned 8-bit image data, the low-order bits of each 8-bit are filled with zeros and expanded into 10-bits for storage, thereby facilitating the support of 8-bit or 10-bit algorithms. So that the subsequent storage to the corresponding buffer.

For the data of the image to be processed with the maximum bit depth, it can be stored according to the expanded storage bit width.

Specifically, the method 100 further includes: obtaining data corresponding to the buffer space for the data of the image to be processed with the maximum bit depth.

For example, according to the foregoing, for 10-bit image data, the 128-bit data received for the first time is temporarily not stored, and the subsequent received 128-bit data is made up into 160-bit and then stored (where the depth is 4). And so on. At the same time, for UV, the UV is interleaved when it comes from the AXI bus. For the convenience of subsequent processing, the UV can be split into the order of UUUUU...VVVV. The bit width of U or V is 160bit, and the depth is 2. The total UV depth is also 4.

For the UV in the 8bit image data, the same low-bit zero-filling is extended to 10bit, and the above-mentioned splicing method is also used, and then placed in the corresponding storage of the UV, and will not be repeated. Only description: As shown in Figure 5, Figure 5 shows the corresponding 160-bit data storage after preprocessing of 8-bit data and 10-bit data. Wherein, each row 502 is composed of 16 10-bit bits 501, which is 16*10=160-bit bits.

In addition, the data of the Y0 row shown in Figure 3 is stored in the internal ping-pong buffer shown in Figure 4 after the above preprocessing is completed, and the corresponding Y0; the data of the Y1 row is stored in the internal after the above preprocessing is completed. The ping-pong buffer corresponds to Y1; the UV data is stored in the corresponding UV in the internal ping-pong buffer after the above preprocessing. As can be seen from the foregoing, for Y0, Y1 or UV, they are all corresponding 160*4 storage structures.

102: Acquire a preset amount of first input data from the data according to the preset data format.

Wherein, the preset number is less than the number of input data conforming to the preset transformation algorithm.

Specifically, acquiring a preset number of first input data from the data according to the preset data format includes: for any cache area, based on the component order of YUV, sequentially acquiring at least two identical corresponding components from any cache area The data is used as the first input data, so that a plurality of corresponding processing results are obtained.

For example, according to the foregoing, when the ping or pong buffer in the ping-pong buffer is full, the subsequent DWT subset module obtains data from the corresponding buffer and starts calculation. The input of Data in (data input) is 160bit, and the data of 20bit bit width is read out at a time, that is, 2 data, each data is 10bit, as the first input data. For example, for Y0, the corresponding two data can be read out at a time, 20bit.

Among them, according to the above method, it can be obtained as shown in Figure 6, the data of Y0601 is read out as two data in turn, which can be two data of idx1 and idx0 in address 0, a total of 64 data, for 64*10=640bit , corresponding to the data mentioned above, each data in160bit. The Y1602 is also similar. For UV603, it is 32 U and 32 V, and the same is also input with two data, so I won't go into details.

103: Based on the obtained first input data, directly multiplex the last set of second input data that conforms to the preset number in the input data of the preset transformation algorithm in the order from back to front, and use the first input data as pending processing Input data.

For example, according to the foregoing, when the DWT subset module in the encoder is reading 1 set of data (that is, two data, 20 bits), it needs to multiplex one data of the previous set, that is, the second input data (except for the beginning part). , that is, outside the first group of data), a total of three data, as input data to be processed, an operation is performed for every three data. This corresponds to the output of 20bit data.

As shown in FIG. 6 , for the first input data idx3 and dix2 in Y0, it is necessary to multiplex dix1 in the previous set of data, so that the final input data to be processed are idx3, dix2 and idx1.

And for the first set of data in the beginning, it can be processed in the following ways:

Specifically, the method 100 further includes: when the first input data belongs to the first group of input data to be processed, performing mirror data supplementation based on the obtained first input data to obtain supplementary data; combining the first input data and the Supplementary data as the first set of input data to be processed.

For example, according to the above, for the first set of data in the beginning, that is, the first input data, such as idx1 and dix0, mirror (automatic mirror data supplementation) needs to be done to obtain supplementary data idx-1, and the first set of pending data is obtained. Process the input data idx1, dix0 and idx-1.

And for the last set of data in the end part, it can be processed in the following ways:

Specifically, the method 100 further includes: after obtaining the last group of first input data from the data and performing calculations based on the obtained corresponding input data to be processed and a preset transformation algorithm, directly complex Use the preset number of second input data in the corresponding input data to be processed; perform mirror data supplementation based on the preset number of second input data to obtain supplementary data; use the preset number of second input data and supplementary data as the final A set of pending input data.

Obtained in a similar manner as described above. For example, according to the method described above, idx61 in the previous group of input data is directly multiplexed on the last group of first input data idx63 and idx62, to obtain corresponding input data idx63, idx62 and idx61 to be processed. And the corresponding calculation results are obtained after the DWT algorithm calculation. Then, the input data idx63, idx62, and idx61 to be processed are directly multiplexed in a back-to-front manner (ie, from left to right) to obtain second input data idx63. Then perform mirror supplementation to obtain supplementary data idx61, idx62, and obtain the last group of pending input data idx61, idx62, and idx63.

104 : Perform calculation according to the input data to be processed and a preset transformation algorithm to obtain a processing result representing the data frequency of the image to be processed.

Wherein, the preset transform algorithm includes discrete wavelet transform algorithm DWT.

Among them, Y needs to perform 2D-DWT, that is, two-dimensional DWT, and UV only needs to perform one-dimensional DWT. According to this method, the above input data to be processed is calculated to obtain a corresponding calculation result. As shown in FIG. 7 , FIG. 7 shows a complete calculation method of 2D-DWT taking the Y component as an example, including transformation parts 701 - 703 . The first transformation part 701 is a one-dimensional transformation part. The second transform part 702 is a 2-dimensional, Haar harr transform part. The fourth transform portion 704 in the third transform portion 703 is a one-dimensional transform portion for L1 in YO. The fifth transform portion 705 is a one-dimensional transform portion for L2 in YO. The sixth transform portion 706 is a one-dimensional transform portion for L3 in YO.

And the 1D DWT for UV is a similar transformation as above. The transformation mode of the Y0 part in 701 , 704 and 705 may be included, and the transformation mode of the corresponding Y1 part in 701 , 704 and 705 is not involved, so it is used as the one-dimensional DWT calculation mode of UV.

Specifically, the calculation is performed according to the input data to be processed and the preset transformation algorithm, including: in the case of performing the one-dimensional transformation algorithm in the preset transformation algorithm, based on the input data to be processed and the high-frequency algorithm in the one-dimensional transformation algorithm, Obtain the current one-dimensional transform high-frequency coefficients representing high-frequency data; based on the current one-dimensional transform high-frequency coefficients, directly multiplex the previous set of one-dimensional transform high-frequency coefficients representing high-frequency data and the low-frequency algorithm in the one-dimensional transform algorithm, The current one-dimensional transform low-frequency coefficients representing the low-frequency data are obtained. The one-dimensional transform high-frequency coefficients of the previous group are obtained from the previous group of input data conforming to the preset transform algorithm and the high-frequency algorithm in the one-dimensional transform algorithm. The one-dimensional transform of the previous group is obtained. High frequency coefficients are not cached.

Among them, the one-dimensional transformation algorithm is one-dimensional DWT.

For example, according to the foregoing, as shown in Figure 8, taking U as an example, the process of calculating the current one-dimensional transform high-frequency coefficient H1:

The corresponding data idx0 and idx1 come in, and the mirror is supplemented with idx-1=idx1, and the three numbers are idx1, idx0 and idx-1. The high frequency algorithm is as follows:

Among them, N=64 of Y0, Y1, N=32 of UV,

Round is the rounding function. The formula 1) is transformed into the following formula 2):

In the above formula 2), n is 15, so the Y (idx31) of H1 is obtained, that is, the idx31 in H1 shown in Figure 8 requires three numbers of x0, x1, x-1, namely idx0, idx1 and idx-1.

The corresponding data idx2 and idx3 come in, and the last data idx1 is used at the same time, and the three numbers use formula 1) At this time, n in formula 1) is 14, and the Y (idx30) of H1 is obtained, which is shown in Figure 8. H1 in idx30.

After the idx31 and idx30 of H1 are obtained, and then according to idx1, L15 in the current one-dimensional transform low-frequency coefficient L1 can be calculated at the same time. That is, the low-frequency algorithm is shown in the following formula 3):

Then Equation 3) changes to Equation 4), where n is 15:

Thus, Y(idxL15) is obtained, which is L15 in L1 shown in FIG. 8 . In this way, the entire calculation process is completely pipelined, and the results of the previous level are given to the next level to participate in the calculation, without additional buffering and fetching in the middle.

Specifically, the calculation is performed according to the input data to be processed and the preset transformation algorithm, including: in the case of performing the two-dimensional transformation algorithm in the preset transformation algorithm, directly multiplexing the corresponding one-dimensional transformation high-frequency coefficients and the two-dimensional transformation algorithm, Obtain the current two-dimensional transform high-frequency coefficients representing high-frequency data; directly multiplex the corresponding one-dimensional transform low-frequency coefficients and two-dimensional transform algorithms to obtain the current two-dimensional transform low-frequency coefficients representing low-frequency data, and the corresponding one-dimensional transform high-frequency coefficients No buffering is performed, and the corresponding one-dimensional transform low-frequency coefficients are not buffered.

Among them, the two-dimensional transformation algorithm refers to the Haar transform in the two-dimensional DWT.

For example, according to the foregoing, for the Y0 component, as shown in Figure 9, Harr H and Harr L are calculated by the following formula 5) Haar transform algorithm:

Among them, the input of formula 5) is H1 of Y0 calculated by the previous stage, H1 of Y1, L1 of Y0, L1 of Y1, and H1 calculated from the previous stage of the Y0 component can refer to Figure 9. It is equivalent to Y1_H1-Y0_H1=Y1_Harr_H; Y1_L1-Y0_L1=Y1_Harr_L1, wherein the L1 calculated from the previous stage of the Y0 component can refer to FIG. 9 . As a result, the Harr two-dimensional transform high-frequency coefficients and low-frequency coefficients of Y1 are obtained.

For the Y1 component, Harr H and Harr L are calculated by the following formula 6) Haar transform algorithm:

Wherein, the formula 6) is H1 of Y0, H1 of Y1, L1 of Y0, and L1 of Y1 calculated from the previous stage. Equivalent to Y0_H1+(Y1_H1-Y0_H1)/2=Y0_Harr_H, Y0_L1+(Y1_L1-Y0_L1)/2=Y0_Harr_L, the result is to obtain the Harr two-dimensional transform high-frequency coefficient and low-frequency coefficient of Y0.

In addition, the method 100 further includes: based on the current two-dimensional transform low-frequency coefficients, directly multiplexing the first two sets of two-dimensional transform low-frequency coefficients representing low-frequency data, and a high-frequency algorithm in the one-dimensional transform algorithm, obtaining a current one representing high-frequency data. The high-frequency coefficients of the two-dimensional transformation are used as the input of the high-frequency algorithm in the one-dimensional transformation algorithm, and the low-frequency coefficients of the two-dimensional transformation are not cached; based on the two-dimensional transformation algorithm The current one-dimensional transform high-frequency coefficients obtained afterward, the one-dimensional transform high-frequency coefficients of the previous group and the low-frequency algorithm in the one-dimensional transform algorithm are directly multiplexed to obtain the current one-dimensional transform low-frequency coefficients representing the low-frequency data.

It is used here to calculate the numerical calculation after Haar transformation as shown in Figure 9, such as H2 and so on. It is calculated by the above formula 1). For the specific process, reference may be made to the foregoing description, which will not be repeated here. Just to explain:

In the subsequent continuous calculation, since the division of the continuous series N will be 32, 16, etc., and so on, then n will also change accordingly.

For example, as shown in Figure 9: H2(16+0=16)=X(31-1=30)-Round((X(31)+X(29))/2) HarrL data is used here , such as L29-31.

After the one-dimensional transform algorithm, the method 100 further includes: based on the obtained current one-dimensional transform low-frequency coefficients, directly multiplexing the first two sets of one-dimensional transform low-frequency coefficients representing low-frequency data, and the high-frequency algorithm in the one-dimensional transform algorithm, obtaining Indicates the current one-dimensional transform high-frequency coefficients of high-frequency data, and the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups can be used as the input of the high-frequency algorithm in the one-dimensional transform algorithm. The one-dimensional transform low-frequency coefficients are not Cache; based on the obtained current one-dimensional transform high-frequency coefficients, the one-dimensional transform high-frequency coefficients of the previous group that are directly multiplexed, and the low-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform low-frequency coefficients representing the low-frequency data are obtained.

It is used here to calculate the numerical calculation after H2 as shown in Figure 9, such as H3 and so on. It is calculated by the above formula 1). For the specific process, reference may be made to the foregoing description, which will not be repeated here. Just to explain:

In the subsequent continuous calculation, since the division of the continuous series N will be 32, 16, etc., and so on, then n will also change accordingly.

After processing the result, the subsequent encoding process can be continued.

Specifically, the method 100 further includes: encoding the image to be processed based on the processing result.

After the DWT calculation is performed on the data of the above image, the corresponding result can be obtained. The output result is dwt-coeff, and the dwt-coeff is expanded to form a bitplane bit plane to complete subsequent image encoding.

Among them, the DWT result is written to the storage through the internal bus line; for Y1, there are at most 2 outputs at a time, corresponding to Harr H and Harr L; for Y0, there are at most 3 outputs at a time, corresponding to Harr H and H2 , H3 or Harr H and H2, H4H or L4. Similarly for UV, there will be at most three valid results output each time. Therefore, it can be written to the internal register file in the form of a bus. Y0: The address line is 6bit, and the data bit width is 12bit*3 for a single output result. Y1: The address line is 6bit, and the data bit width is 12bit*2 for a single output result. UV: address line 5bit, data bit width is 12bit*3 for a single output result. The entire process does not require additional caching, and the data is pipline executed to improve efficiency.

In addition, according to your own needs, you can change the number of feeds at the beginning of the pipeline from 3 to 5, and the efficiency is doubled. If performance is limited, you can take measures such as adding some resources.

After the encoding is completed, the encoded image stream can be subjected to subsequent processing, such as transmission, storage, and decoding.

Specifically, the method 100 further includes: storing the encoded image to be processed in a corresponding storage area for transmission or decoding.

As can be seen from the foregoing, as shown in FIG. 2 , the FBC encoder 202 can send the encoded code stream, that is, the image code stream, to the DDR memory 203 for storage, waiting for decoding, or can be transmitted to other devices. During decoding, the encoded code stream in the DDR memory 203 can be decoded by the FBC decoder 204 .

FIG. 10 is a schematic structural diagram of a data processing apparatus provided by an embodiment of the present invention; the apparatus 1000 can be applied to a flying device, such as an unmanned aerial vehicle, more specifically an FBC encoder, and the apparatus 1000 can execute the above-mentioned data processing method. The apparatus 1000 includes: an acquisition module 1001 , a multiplexing module 1002 and a processing module 10003 . The functions of each module are described in detail below:

The obtaining module 1001 is configured to obtain the to-be-processed image in a preset data format, and store the data of the to-be-processed image according to the preset storage format.

The obtaining module 1001 is configured to obtain a preset amount of first input data from the data according to a preset data format.

Wherein, the preset number is less than the number of input data conforming to the preset transformation algorithm.

The multiplexing module 1002 is configured to, based on the acquired first input data, directly multiplex the last set of second input data in accordance with the preset number of input data of the preset transformation algorithm in the order from back to front, and combine them with the first input data. Input data as pending input data.

The processing module 1003 is configured to perform calculation according to the input data to be processed and a preset transformation algorithm to obtain a processing result representing the data frequency of the image to be processed.

In addition, the apparatus 1000 further includes: an encoding module, configured to encode the image to be processed based on the processing result.

In addition, the processing module 1003 includes: a first determining unit, configured to obtain a high-frequency representation based on the input data to be processed and a high-frequency algorithm in the one-dimensional The current one-dimensional transform high-frequency coefficients of the high-frequency data; the second determination unit is used for directly multiplexing the previous one-dimensional transform high-frequency coefficients representing the high-frequency data based on the current one-dimensional transform high-frequency coefficients and the one-dimensional transform algorithm. The low-frequency algorithm is used to obtain the current one-dimensional transform low-frequency coefficients representing the low-frequency data. The one-dimensional transform high-frequency coefficients of the previous group are obtained from the input data of the previous group conforming to the preset transform algorithm and the high-frequency algorithm in the one-dimensional transform algorithm. The high-frequency coefficients of the 1D transform are not cached.

Specifically, the processing module 1003 includes: a third determining unit, configured to directly multiplex the corresponding one-dimensional transform high-frequency coefficients and the two-dimensional transform algorithm in the case of performing the two-dimensional transform algorithm in the preset transform algorithm to obtain the representation The current two-dimensional transform high-frequency coefficients of the high-frequency data; the fourth determining unit is used to directly multiplex the corresponding one-dimensional transform low-frequency coefficients and the two-dimensional transform algorithm to obtain the current two-dimensional transform low-frequency coefficients representing the low-frequency data. The high-frequency coefficients of the one-dimensional transform are not cached, and the corresponding low-frequency coefficients of the one-dimensional transform are not cached.

In addition, the processing module 1003 is also used for: based on the current two-dimensional transform low-frequency coefficients, directly multiplexing the first two sets of two-dimensional transform low-frequency coefficients representing low-frequency data, and the high-frequency algorithm in the one-dimensional transform algorithm, to obtain the high-frequency data representing the high-frequency data. The current one-dimensional transform high-frequency coefficients, and the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups are used as the input of the high-frequency algorithm in the one-dimensional transform algorithm, and the two-dimensional transform low-frequency coefficients are not cached; based on the two-dimensional transform The current one-dimensional transform high-frequency coefficients obtained after the transform algorithm, the one-dimensional transform high-frequency coefficients of the previous group and the low-frequency algorithm in the one-dimensional transform algorithm are directly multiplexed to obtain the current one-dimensional transform low-frequency coefficients representing the low-frequency data.

In addition, after the one-dimensional transform algorithm, the processing module 1003 is further configured to: based on the obtained current one-dimensional transform low-frequency coefficients, directly multiplex the first two sets of one-dimensional transform low-frequency coefficients representing low-frequency data, and the high-frequency coefficients in the one-dimensional transform algorithm. frequency algorithm to obtain the current one-dimensional transform high-frequency coefficients representing high-frequency data, and the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups can be used as the input of the high-frequency algorithm in the one-dimensional transform algorithm. The transformed low-frequency coefficients are not cached; based on the obtained current one-dimensional transform high-frequency coefficients, the one-dimensional transform high-frequency coefficients of the previous group that are directly multiplexed, and the low-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform representing the low-frequency data is obtained. low frequency coefficients.

Wherein, the preset transform algorithm includes discrete wavelet transform algorithm.

In addition, the processing module 1003 is further configured to: when the first input data belongs to the first group of input data to be processed, perform mirror data supplementation based on the obtained first input data to obtain supplementary data; and the supplementary data as the first set of input data to be processed.

In addition, the multiplexing module 1002 is also used for: after obtaining the last group of first input data from the data and calculating based on the corresponding input data to be processed and the preset transformation algorithm, follow the sequence from back to front. Directly multiplexing a preset number of second input data in the corresponding input data to be processed; the device 1000 further includes a determination module for performing mirror data supplementation based on the preset number of second input data to obtain supplementary data; The set number of second input data and supplementary data are used as the final set of pending input data.

Specifically, the acquisition module 1001 includes: an acquisition unit for acquiring an image in a YUV preset data format as an image to be processed; a cache unit for sequentially caching the first group of YUV component data to the first group of YUV component data according to the YUV component order A buffer space corresponding to the buffer area; the buffer unit is used for sequentially buffering the YUV component data of the second group to the corresponding buffer space in the second buffer area after buffering the YUV component data of the first group. After the first group of YUV component data is cached, the acquiring module 1001 performs the step of acquiring a preset number of first input data from the data according to the preset data format to obtain the processing result, and the cache unit is used to continue to cache the corresponding group The YUV component data is stored in the first buffer area, until the YUV component data is buffered; after the second group of YUV component data is buffered, the acquisition module 1001 executes according to the preset data format, from the data to obtain a preset number of first input data step, so that when the processing result is reached, the buffer unit is used to continue buffering the YUV component data of the corresponding group to the second buffer area until the YUV component data is buffered.

Specifically, the acquiring module 1001 is configured to: for any buffer area, based on the component order of YUV, sequentially acquire at least two identical corresponding component data from any buffer area as the first input data, so that a plurality of corresponding component data are obtained. process result.

The buffer space buffers data of at least two images to be processed with different bit depths.

In addition, the apparatus 1000 further includes: an expansion module, configured to expand the corresponding original preset cache space in any cache area based on the data of the image to be processed with the largest bit depth in at least two different bit depths, and generate a corresponding In the buffer space, the original preset buffer space buffers the data of the image to be processed with the smallest bit depth among at least two different bit depths.

In addition, the device 1000 also includes: an expansion module for performing data expansion on the data of the to-be-processed image with a non-maximum bit-depth in at least two different bit depths, to obtain the data of the to-be-processed image with the maximum bit depth, and buffering the data into the buffer space.

In addition, the obtaining module 1001 is further configured to obtain data corresponding to the buffer space for the data of the image to be processed with the maximum bit depth.

In addition, the apparatus 1000 further includes: a storage module for storing the encoded image to be processed in a corresponding storage area for transmission or decoding.

It should be noted that, for parts that are not described in detail in this embodiment, reference may be made to the relevant descriptions of the embodiments shown in the image processing method above, and details are not described herein again.

In a possible design, the structure of the data processing apparatus 1000 shown in FIG. 10 may be implemented as an electronic device, and the electronic device may be a data processing device. As shown in FIG. 11 , the data processing device 1100 may include: one or more processors 1101 and one or more memories 1102 . Wherein, the memory 1102 is used for storing a program that supports the electronic device to execute the data processing method provided in the embodiments shown in FIG. 1 to FIG. 9 . The processor 1101 is configured to execute programs stored in the memory 1102 . Specifically, the program includes one or more computer instructions, wherein the one or more computer instructions can realize the following steps when executed by the processor 1101:

The computer program stored in the memory 1102 is run to realize: obtaining the image to be processed in a preset data format, and storing the data of the image to be processed according to the preset storage format; obtaining a preset number of first inputs from the data according to the preset data format Data, the preset number is less than the number of input data conforming to the preset transformation algorithm; based on the obtained first input data, the presets in the previous set of input data conforming to the preset transformation algorithm are directly reused in the order from back to front The number of second input data is combined with the first input data as input data to be processed; calculation is performed according to the input data to be processed and a preset transformation algorithm, and a processing result representing the data frequency of the image to be processed is obtained.

In addition, the processor 1101 is further configured to: encode the image to be processed based on the processing result.

Specifically, the processor 1101 is specifically configured to: in the case of performing the one-dimensional transformation algorithm in the preset transformation algorithm, based on the input data to be processed and the high-frequency algorithm in the one-dimensional transformation algorithm, obtain the current value representing the high-frequency data. One-dimensional transform high-frequency coefficients; based on the current one-dimensional transform high-frequency coefficients, directly multiplexing the previous set of one-dimensional transform high-frequency coefficients representing high-frequency data, and the low-frequency algorithm in the one-dimensional transform algorithm, obtain the current one-dimensional transform representing low-frequency data. The low-frequency coefficients of the one-dimensional transformation are obtained from the input data of the previous group conforming to the preset transformation algorithm and the high-frequency algorithm in the one-dimensional transformation algorithm, and the high-frequency coefficients of the one-dimensional transformation of the previous group are not cached.

Specifically, the processor 1101 is specifically configured to: in the case of performing the two-dimensional transform algorithm in the preset transform algorithm, directly multiplex the corresponding one-dimensional transform high-frequency coefficients and the two-dimensional transform algorithm to obtain high-frequency data representing high-frequency data. The current two-dimensional transform high-frequency coefficients are directly multiplexed with the corresponding one-dimensional transform low-frequency coefficients and the two-dimensional transform algorithm to obtain the current two-dimensional transform low-frequency coefficients representing low-frequency data, and the corresponding one-dimensional transform high-frequency coefficients are not cached , the corresponding one-dimensional transform low-frequency coefficients are not cached.

In addition, the processor 1101 is further configured to: based on the current two-dimensional transform low-frequency coefficients, directly multiplexing the first two sets of two-dimensional transform low-frequency coefficients representing low-frequency data, and the high-frequency algorithm in the one-dimensional transform algorithm, to obtain the high-frequency data representing the high-frequency data. The current one-dimensional transform high-frequency coefficients, and the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups are used as the input of the high-frequency algorithm in the one-dimensional transform algorithm, and the two-dimensional transform low-frequency coefficients are not cached; based on the two-dimensional transform The current one-dimensional transform high-frequency coefficients obtained after the transform algorithm, the one-dimensional transform high-frequency coefficients of the previous group and the low-frequency algorithm in the one-dimensional transform algorithm are directly multiplexed to obtain the current one-dimensional transform low-frequency coefficients representing the low-frequency data.

In addition, after the one-dimensional transform algorithm, the processor 1101 is further configured to: based on the obtained current one-dimensional transform low-frequency coefficients, directly multiplex the first two sets of one-dimensional transform low-frequency coefficients representing low-frequency data, and the high-frequency coefficients in the one-dimensional transform algorithm. frequency algorithm to obtain the current one-dimensional transform high-frequency coefficients representing high-frequency data, and the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups can be used as the input of the high-frequency algorithm in the one-dimensional transform algorithm. The transformed low-frequency coefficients are not cached; based on the obtained current one-dimensional transform high-frequency coefficients, the one-dimensional transform high-frequency coefficients of the previous group that are directly multiplexed, and the low-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform representing the low-frequency data is obtained. low frequency coefficients.

Wherein, the preset transform algorithm includes discrete wavelet transform algorithm.

In addition, the processor 1101 is further configured to: when the first input data belongs to the first group of input data to be processed, perform mirror data supplementation based on the acquired first input data to obtain supplementary data; and supplementary data as the first set of pending input data.

In addition, the processor 1101 is further configured to: when the last group of first input data is obtained from the data and the calculation is performed based on the obtained corresponding input data to be processed and the preset transformation algorithm, directly follow the sequence from back to front. Multiplexing the preset number of second input data in the corresponding input data to be processed; performing mirror data supplementation based on the preset number of second input data to obtain supplementary data; using the preset number of second input data and supplementary data as The last set of pending input data.

In addition, the processor 1101 is specifically used to: obtain an image in a YUV preset data format as an image to be processed; according to the component order of YUV, sequentially cache the first group of YUV component data in the corresponding cache space in the first cache area; After buffering the first group of YUV component data, buffer the second group of YUV component data in the corresponding buffer space in the second buffer area in turn; The step of obtaining a preset number of first input data in the data, so that to the processing result, continue to cache the YUV component data of the corresponding group to the first cache area until the YUV component data is cached; after the second group of YUV component data is cached After the data, perform the step of obtaining a preset number of first input data from the data according to the preset data format to obtain the processing result, and continue to cache the YUV component data of the corresponding group to the second cache area until the YUV is cached. component data.

Specifically, the processor 1101 is specifically configured to: for any cache area, based on the component order of YUV, sequentially acquire at least two identical corresponding component data from any cache area as the first input data, so that multiple corresponding component data are obtained. processing result.

The buffer space buffers data of at least two images to be processed with different bit depths.

In addition, the processor 1101 is further configured to: expand the corresponding original preset cache space in any cache area based on the data of the image to be processed with the largest bit depth among the at least two different bit depths, and generate the corresponding cache space, The original preset buffer space buffers the data of the image to be processed with the smallest bit depth among at least two different bit depths.

In addition, the processor 1101 is further configured to: perform data expansion on the data of the to-be-processed image with a non-maximum bit-depth among at least two different bit-depths, to obtain the data of the to-be-processed image with the maximum bit depth, so as to be cached in the cache space .

In addition, the processor 1101 is further configured to obtain data corresponding to the cache space for the data of the image to be processed with the maximum bit depth.

In addition, the processor 1101 is further configured to: store the encoded image to be processed in a corresponding storage area for transmission or decoding.

In addition, an embodiment of the present invention provides a computer-readable storage medium, where the storage medium is a computer-readable storage medium, and program instructions are stored in the computer-readable storage medium, and the program instructions are used to implement the above-mentioned methods in FIGS. 1 to 9 . .

A chip, such as an SOC, provided by an embodiment of the present invention includes: the above-mentioned device.

An embodiment of the present invention provides an unmanned aerial vehicle; specifically, the unmanned aerial vehicle includes: a body and a data processing device shown in FIG. 11 , and the data processing device is arranged on the body.

The technical solutions and technical features in the above embodiments can be used alone or combined in the case of conflict with the present invention, as long as they do not exceed the cognitive scope of those skilled in the art, they all belong to the equivalent embodiments within the protection scope of the present application .

In the several embodiments provided by the present invention, it should be understood that the disclosed related detection apparatus (eg, IMU) and method may be implemented in other manners. For example, the embodiments of the remote control device described above are only illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components. May be combined or may be integrated into another system, or some features may be omitted, or not implemented. On the other hand, the shown or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, and the indirect coupling or communication connection of the remote control device or unit may be in electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution in this embodiment.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium , including several instructions for causing a computer processor (processor) to perform all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes.

The above descriptions are only the embodiments of the present invention, and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the description and drawings of the present invention, or directly or indirectly applied to other related technologies Fields are similarly included in the scope of patent protection of the present invention.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, but not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: The technical solutions described in the foregoing embodiments can still be modified, or some or all of the technical features thereof can be equivalently replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the technical solutions of the embodiments of the present invention. scope.

Claims (35)

1. A data processing method, comprising:

   Acquiring a to-be-processed image in a preset data format, and storing the data of the to-be-processed image according to a preset storage format;

   According to the preset data format, a preset number of first input data is obtained from the data, and the preset number is less than the number of input data conforming to the preset transformation algorithm;

   Based on the acquired first input data, directly multiplex the last set of second input data that conforms to the preset number in the input data of the preset transformation algorithm in the order from back to front, and combine them with the first input data. data as input data to be processed;

   The calculation is performed according to the input data to be processed and the preset transformation algorithm to obtain a processing result representing the data frequency of the image to be processed.
2. The method according to claim 1, wherein the method further comprises:

   Based on the processing result, the image to be processed is encoded.
3. The method according to claim 1, wherein the calculating according to the input data to be processed and the preset transformation algorithm comprises:

   In the case of performing the one-dimensional transformation algorithm in the preset transformation algorithm, based on the input data to be processed and the high-frequency algorithm in the one-dimensional transformation algorithm, a current one-dimensional transformation high-frequency representation representing the high-frequency data is obtained. coefficient;

   Based on the current one-dimensional transform high-frequency coefficients, directly multiplexing the previous set of one-dimensional transform high-frequency coefficients representing high-frequency data, and the low-frequency algorithm in the one-dimensional transform algorithm, obtain the current one-dimensional transform low-frequency coefficient representing the low-frequency data. Coefficients, the one-dimensional transform high-frequency coefficients of the previous group are obtained by the previous group conforming to the input data of the preset transform algorithm and the high-frequency algorithm in the one-dimensional transform algorithm, and the one-dimensional transform high-frequency coefficients of the previous group are not. cache.
4. The method according to claim 1 or 3, wherein the calculating according to the input data to be processed and the preset transformation algorithm comprises:

   In the case of performing the two-dimensional transform algorithm in the preset transform algorithm, directly multiplexing the corresponding one-dimensional transform high-frequency coefficients and the two-dimensional transform algorithm to obtain the current two-dimensional transform high-frequency coefficients representing high-frequency data;

   Directly multiplex the corresponding one-dimensional transform low-frequency coefficients and the two-dimensional transform algorithm to obtain the current two-dimensional transform low-frequency coefficients representing low-frequency data, the corresponding one-dimensional transform high-frequency coefficients are not cached, and the corresponding one-dimensional transform low-frequency coefficients Coefficients are not cached.
5. The method according to claim 4, wherein the method further comprises:

   Based on the current two-dimensional transform low-frequency coefficients, directly multiplexing the first two sets of two-dimensional transform low-frequency coefficients representing low-frequency data, and a high-frequency algorithm in the one-dimensional transform algorithm, obtain the current one-dimensional transform high-frequency coefficients representing high-frequency data, And the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups are used as the input of the high-frequency algorithm in the one-dimensional transform algorithm, and the two-dimensional transform low-frequency coefficients are not cached;

   Based on the current one-dimensional transform high-frequency coefficients obtained after the two-dimensional transform algorithm, directly multiplexing the one-dimensional transform high-frequency coefficients of the previous group, and the low-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform low-frequency coefficients representing the low-frequency data are obtained. .
6. The method according to claim 3, wherein after the one-dimensional transformation algorithm, the method further comprises:

   Based on the obtained current one-dimensional transform low-frequency coefficients, the first two groups of one-dimensional transform low-frequency coefficients representing low-frequency data, and the high-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform high-frequency coefficients representing high-frequency data are obtained, And the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups can be used as the input of the high-frequency algorithm in the one-dimensional transform algorithm, and the one-dimensional transform low-frequency coefficients are not cached;

   Based on the obtained current one-dimensional transform high-frequency coefficients, directly multiplexing the one-dimensional transform high-frequency coefficients of the previous group, and the low-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform low-frequency coefficients representing the low-frequency data are obtained.
7. The method according to claim 1, wherein the preset transform algorithm comprises a discrete wavelet transform algorithm.
8. The method according to claim 1, wherein the method further comprises:

   When the first input data belongs to the first group of input data to be processed, performing mirror data supplementation based on the acquired first input data to obtain supplementary data;

   The first input data and the supplementary data are used as the first group of input data to be processed.
9. The method according to claim 1, wherein the method further comprises:

   When the last group of first input data is obtained from the data and calculated based on the corresponding input data to be processed and the preset transformation algorithm, the corresponding a preset number of second input data in the input data to be processed;

   Perform mirror data supplementation based on the preset number of second input data to obtain supplementary data;

   The preset number of second input data and the supplementary data are used as the input data to be processed in the last group.
10. The method according to claim 1, wherein obtaining the image to be processed in the preset data format, and storing the data of the image to be processed according to the preset storage format, comprising:

    Obtain an image in the YUV preset data format as an image to be processed;

    According to the component order of YUV, the first group of YUV component data is sequentially cached in the corresponding buffer space in the first cache area; after the first group of YUV component data is cached, the second group of YUV component data is sequentially cached in the second cache. The corresponding cache space in the region;

    The method further includes: after buffering the first group of YUV component data, performing the step of obtaining a preset number of first input data from the data according to the preset data format, to obtain the processing As a result, continue to buffer the YUV component data of the corresponding group to the first buffer area until the YUV component data is buffered;

    After the second set of YUV component data is cached, the step of obtaining a preset number of first input data from the data according to the preset data format is performed, so that the processing result is continued to be cached The YUV component data of the corresponding group is stored in the second buffer area until the YUV component data is buffered.
11. The method according to claim 10, wherein the obtaining a preset number of first input data from the data according to the preset data format comprises:

    For any buffer area, based on the component order of YUV, at least two identical corresponding component data are sequentially acquired from any buffer area as the first input data, so as to obtain multiple corresponding processing results.
12. The method according to claim 10, wherein the buffer space buffers data of images to be processed with at least two different bit depths.
13. The method of claim 10, wherein the method further comprises:

    Based on the data of the image to be processed with the largest bit depth among at least two different bit depths, the corresponding original preset cache space in any cache area is expanded to generate the corresponding cache space, and the original preset cache space is cached The data of the image to be processed with the smallest bit depth among at least two different bit depths.
14. The method according to claim 12 or 13, wherein the method further comprises:

    Perform data expansion on the data of the to-be-processed image with a non-maximum bit-depth among the at least two different bit depths, to obtain the data of the to-be-processed image with the maximum bit depth, so as to be cached in the cache space.
15. The method according to claim 12 or 13, wherein the method further comprises:

    For the data of the image to be processed with the maximum bit depth, data corresponding to the cache space is acquired.
16. The method according to claim 2, wherein the method further comprises:

    Store the encoded image to be processed in the corresponding storage area for transmission or decoding.
17. A data processing device, comprising: a memory, a processor;

    the memory for storing computer programs;

    The processor invokes the computer program to implement the following steps:

    Acquiring a to-be-processed image in a preset data format, and storing the data of the to-be-processed image according to a preset storage format;

    According to the preset data format, a preset number of first input data is obtained from the data, and the preset number is less than the number of input data conforming to the preset transformation algorithm;

    Based on the acquired first input data, directly multiplex the last set of second input data that conforms to the preset number in the input data of the preset transformation algorithm in the order from back to front, and combine them with the first input data. data as input data to be processed;

    The calculation is performed according to the input data to be processed and the preset transformation algorithm to obtain a processing result representing the data frequency of the image to be processed.
18. The device of claim 17, wherein the processor is further configured to:

    Based on the processing result, the image to be processed is encoded.
19. The device according to claim 17, wherein the processor is specifically configured to:

    In the case of performing the one-dimensional transformation algorithm in the preset transformation algorithm, based on the input data to be processed and the high-frequency algorithm in the one-dimensional transformation algorithm, a current one-dimensional transformation high-frequency representation representing the high-frequency data is obtained. coefficient;

    Based on the current one-dimensional transform high-frequency coefficients, directly multiplexing the previous set of one-dimensional transform high-frequency coefficients representing high-frequency data, and the low-frequency algorithm in the one-dimensional transform algorithm, obtain the current one-dimensional transform low-frequency coefficient representing the low-frequency data. Coefficients, the one-dimensional transform high-frequency coefficients of the previous group are obtained by the previous group conforming to the input data of the preset transform algorithm and the high-frequency algorithm in the one-dimensional transform algorithm, and the one-dimensional transform high-frequency coefficients of the previous group are not. cache.
20. The device according to claim 17 or 19, wherein the processor is specifically configured to:

    In the case of performing the two-dimensional transform algorithm in the preset transform algorithm, directly multiplexing the corresponding one-dimensional transform high-frequency coefficients and the two-dimensional transform algorithm to obtain the current two-dimensional transform high-frequency coefficients representing high-frequency data;

    Directly multiplex the corresponding one-dimensional transform low-frequency coefficients and the two-dimensional transform algorithm to obtain the current two-dimensional transform low-frequency coefficients representing low-frequency data, the corresponding one-dimensional transform high-frequency coefficients are not cached, and the corresponding one-dimensional transform low-frequency coefficients Coefficients are not cached.
21. The device according to claim 20, wherein the processor is further configured to:

    Based on the current two-dimensional transform low-frequency coefficients, directly multiplexing the first two sets of two-dimensional transform low-frequency coefficients representing low-frequency data, and a high-frequency algorithm in the one-dimensional transform algorithm, obtain the current one-dimensional transform high-frequency coefficients representing high-frequency data, And the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups are used as the input of the high-frequency algorithm in the one-dimensional transform algorithm, and the two-dimensional transform low-frequency coefficients are not cached;

    Based on the current one-dimensional transform high-frequency coefficients obtained after the two-dimensional transform algorithm, directly multiplexing the one-dimensional transform high-frequency coefficients of the previous group, and the low-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform low-frequency coefficients representing the low-frequency data are obtained. .
22. The device according to claim 19, wherein after the one-dimensional transformation algorithm, the processor is further configured to:

    Based on the obtained current one-dimensional transform low-frequency coefficients, the first two groups of one-dimensional transform low-frequency coefficients representing low-frequency data, and the high-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform high-frequency coefficients representing high-frequency data are obtained, And the current two-dimensional transform low-frequency coefficients and the two-dimensional transform low-frequency coefficients of the first two groups can be used as the input of the high-frequency algorithm in the one-dimensional transform algorithm, and the one-dimensional transform low-frequency coefficients are not cached;

    Based on the obtained current one-dimensional transform high-frequency coefficients, directly multiplexing the one-dimensional transform high-frequency coefficients of the previous group, and the low-frequency algorithm in the one-dimensional transform algorithm, the current one-dimensional transform low-frequency coefficients representing the low-frequency data are obtained.
23. The device according to claim 17, wherein the preset transform algorithm comprises a discrete wavelet transform algorithm.
24. The device of claim 17, wherein the processor is further configured to:

    When the first input data belongs to the first group of input data to be processed, performing mirror data supplementation based on the acquired first input data to obtain supplementary data;

    The first input data and the supplementary data are used as the first group of input data to be processed.
25. The device of claim 17, wherein the processor is further configured to:

    When the last group of first input data is obtained from the data and calculated based on the corresponding input data to be processed and the preset transformation algorithm, the corresponding a preset number of second input data in the input data to be processed;

    Perform mirror data supplementation based on the preset number of second input data to obtain supplementary data;

    The preset number of second input data and the supplementary data are used as the input data to be processed in the last group.
26. The device according to claim 17, wherein the processor is specifically configured to:

    Obtain an image in the YUV preset data format as an image to be processed;

    According to the component order of YUV, the first group of YUV component data is sequentially cached in the corresponding buffer space in the first cache area; after the first group of YUV component data is cached, the second group of YUV component data is sequentially cached in the second cache. The corresponding cache space in the region;

    After the first set of YUV component data is cached, the step of obtaining a preset number of first input data from the data according to the preset data format is performed, so that the processing result is continued to be cached The YUV component data of the corresponding group is sent to the first buffer area, until the YUV component data is buffered;

    After the second group of YUV component data is cached, the step of obtaining a preset amount of first input data from the data according to the preset data format is performed to obtain the processing result, and the corresponding group continues to be cached The YUV component data is stored in the second buffer area until the YUV component data is buffered.
27. The device according to claim 26, wherein the processor is specifically configured to:

    For any buffer area, based on the component order of YUV, at least two identical corresponding component data are sequentially acquired from any buffer area as the first input data, so as to obtain multiple corresponding processing results.
28. The device according to claim 26, wherein the buffer space buffers data of images to be processed with at least two different bit depths.
29. The device of claim 26, wherein the processor is further configured to:

    Based on the data of the image to be processed with the largest bit depth among at least two different bit depths, the corresponding original preset cache space in any cache area is expanded to generate the corresponding cache space, and the original preset cache space is cached The data of the image to be processed with the smallest bit depth among at least two different bit depths.
30. The device according to claim 28 or 29, wherein the processor is further configured to:

    Perform data expansion on the data of the to-be-processed image with a non-maximum bit-depth among the at least two different bit depths, to obtain the data of the to-be-processed image with the maximum bit depth, so as to be cached in the cache space.
31. The device according to claim 28 or 29, wherein the processor is further configured to:

    For the data of the image to be processed with the maximum bit depth, data corresponding to the cache space is acquired.
32. The device according to claim 18, wherein the processor is further configured to: store the encoded image to be processed in a corresponding storage area for transmission or decoding.
33. A chip, characterized by comprising: the device according to any one of claims 17-32.
34. An unmanned aerial vehicle, characterized by comprising: a body and the device according to any one of claims 17-32.
35. A computer-readable storage medium, characterized in that the storage medium is a computer-readable storage medium, and program instructions are stored in the computer-readable storage medium, and the program instructions are used to implement any one of claims 1-16 The data processing method described in item.

Images