CN115456858A - Image processing method, image processing device, computer equipment and computer readable storage medium - Google Patents
Image processing method, image processing device, computer equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN115456858A CN115456858A CN202211127995.9A CN202211127995A CN115456858A CN 115456858 A CN115456858 A CN 115456858A CN 202211127995 A CN202211127995 A CN 202211127995A CN 115456858 A CN115456858 A CN 115456858A
- Authority
- CN
- China
- Prior art keywords
- pixel
- image
- unit
- shuffling
- processed
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000012545 processing Methods 0.000 title claims abstract description 41
- 238000003672 processing method Methods 0.000 title claims abstract description 20
- 238000010586 diagram Methods 0.000 claims abstract description 38
- 238000000034 method Methods 0.000 claims abstract description 32
- 238000004590 computer program Methods 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000004891 communication Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 6
- 239000011159 matrix material Substances 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000012546 transfer Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000013139 quantization Methods 0.000 description 2
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 229910021389 graphene Inorganic materials 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008707 rearrangement Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/20—Processor architectures; Processor configuration, e.g. pipelining
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/60—Memory management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4046—Scaling of whole images or parts thereof, e.g. expanding or contracting using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
- G06T3/4053—Scaling of whole images or parts thereof, e.g. expanding or contracting based on super-resolution, i.e. the output image resolution being higher than the sensor resolution
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Image Processing (AREA)
Abstract
The application provides an image processing method, an image processing device, computer equipment and a computer readable storage medium, and relates to the technical field of computers. The method comprises the following steps: reading image pixel data corresponding to pixel units in an image to be processed from an external storage device; performing pixel shuffling on image pixel data corresponding to the pixel units to obtain characteristic diagrams corresponding to the pixel units, and outputting the characteristic diagrams corresponding to the pixel units to an external storage device; taking a pixel unit formed by the next pixel point of each pixel point in the pixel units as a pixel unit; returning to the step of reading image pixel data corresponding to the pixel units in the image to be processed from an external storage device until feature maps corresponding to all the pixel units in the image to be processed are obtained; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed. By adopting the method and the device, the pixel shuffling efficiency can be improved on a computing device with limited computing resources.
Description
Technical Field
The present application relates to the field of computer technologies, and in particular, to an image processing method and apparatus, a computer device, and a computer-readable storage medium.
Background
With the development of artificial intelligence technology, deep neural networks have appeared, and in the structure of the deep neural networks, it is a very common operation to amplify feature maps. The pixel shuffling is called as a depth-to-space operator as an implementation mode for realizing the sub-pixel convolution, and the function of the sub-pixel convolution is realized by rearranging matrix elements in the implementation process.
However, the existing pixel shuffling or depth to space block operators directly perform corresponding dimension conversion on the complete matrix data or image data. Therefore, after the complete matrix or image data is taken out from the external storage device, a large amount of data operation is also required to implement pixel shuffling, i.e. data rearrangement of the complete matrix or image, which is not favorable for the deployment of a neural network on a computing device with limited computing resources.
Disclosure of Invention
The application provides an image processing method, an image processing device, computer equipment and a computer readable storage medium, which can improve the pixel shuffling efficiency on computing equipment with limited computing resources.
In a first aspect, the present application provides an image processing method applied to a pixel shuffling unit connected to an external storage device, where the external storage device stores an image to be processed, the method including:
reading image pixel data corresponding to pixel units in an image to be processed from an external storage device; wherein, the pixel unit is composed of at least one pixel point;
performing pixel shuffling on image pixel data corresponding to the pixel units to obtain characteristic diagrams corresponding to the pixel units, and outputting the characteristic diagrams corresponding to the pixel units to an external storage device; pixel shuffling is to perform convolution processing on image pixel data;
taking a pixel unit consisting of a next pixel point of each pixel point in the pixel units as a pixel unit;
returning to the step of reading image pixel data corresponding to the pixel units in the image to be processed from the external storage device until feature maps corresponding to all the pixel units in the image to be processed are obtained; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed.
In a second aspect, the present application further provides an image processing apparatus applied to a pixel shuffling unit connected to an external storage device, where the external storage device stores an image to be processed, the apparatus comprising:
the reading module is used for reading image pixel data corresponding to the pixel units in the image to be processed from the external storage equipment; wherein, the pixel unit is composed of at least one pixel point;
the shuffling module is used for carrying out pixel shuffling on image pixel data corresponding to the pixel unit to obtain a characteristic diagram corresponding to the pixel unit and outputting the characteristic diagram corresponding to the pixel unit to an external storage device; pixel shuffling is the convolution processing of image pixel data;
the replacing module is used for taking a pixel unit formed by the next pixel point of each pixel point in the pixel unit as the pixel unit;
the circulation module is used for returning to the step of reading image pixel data corresponding to the pixel units in the image to be processed from the external storage device until obtaining the feature maps corresponding to all the pixel units in the image to be processed; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed.
In a third aspect, the present application further provides a computer device, where the computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the image processing method when executing the computer program.
In a fourth aspect, the present application further provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the image processing method described above.
In a fifth aspect, the present application is added to provide a computer program product comprising a computer program that, when executed by a processor, performs the steps of the image processing method described above.
The method for arranging the pixel shuffling unit in the server can directly shuffle the pixels of any pixel unit of the picture to be processed, and by using the newly proposed address operation, the pixel units which do not need to be processed can be temporarily stored in the external storage device, and only the pixel units which are being processed need to be taken out of the external storage device, so that volatile storage resources of heterogeneous computing devices are saved. Meanwhile, a large amount of shifting or similar operations are not needed to be carried out on data in volatile storage resources of the heterogeneous computing equipment, pixel shuffling can be completed directly at the stage of data taking out, and the data processing efficiency is greatly improved.
Drawings
Fig. 1 is an application environment diagram of an image processing method according to an embodiment of the present application;
fig. 2 is a schematic flowchart of an image processing method according to an embodiment of the present application;
fig. 3 is a schematic flowchart of a method for obtaining a feature map corresponding to a pixel unit according to an embodiment of the present disclosure;
fig. 4 is a schematic flowchart of a method for obtaining image pixel data corresponding to a pixel point according to an embodiment of the present application;
fig. 5 is a schematic flowchart of a method for reading image pixel data corresponding to a pixel unit according to an embodiment of the present disclosure;
fig. 6 is a schematic flowchart of a method for obtaining a pixel address corresponding to any pixel point according to an embodiment of the present disclosure;
fig. 7 is a schematic flowchart of a multithread calculation determination method according to an embodiment of the present application;
fig. 8 is a block diagram of an image processing apparatus according to an embodiment of the present application;
fig. 9 is an internal structural diagram of a computer device according to an embodiment of the present application;
fig. 10 is a block diagram of a computer-readable storage medium according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of and not restrictive on the broad application.
The image processing method provided by the embodiment of the application can be applied to the application environment shown in fig. 1. Wherein, the external storage device 102 acquires data; the server 104 receives data from the external storage device 102 in response to the instruction, and calculates the acquired data; the server 104 transmits the calculation result of the data back to the external storage device 102 and is displayed by the external storage device 102. The external storage device 102 communicates with the server 104 through a communication network. The server 104 reads image pixel data corresponding to pixel units in the image to be processed from the external storage device 102; wherein, the pixel unit is composed of at least one pixel point; performing pixel shuffling on image pixel data corresponding to the pixel units to obtain characteristic diagrams corresponding to the pixel units, and outputting the characteristic diagrams corresponding to the pixel units to an external storage device; pixel shuffling is to perform convolution processing on image pixel data; taking a pixel unit consisting of a next pixel point of each pixel point in the pixel units as a pixel unit; returning to the step of reading image pixel data corresponding to the pixel units in the image to be processed from the external storage device until obtaining the feature maps corresponding to all the pixel units in the image to be processed; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed. The external storage device 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, internet of things devices, and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart car-mounted devices, and the like. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, and the like. The server 104 may be implemented as a stand-alone server or as a server cluster comprised of multiple servers.
In some embodiments, as shown in fig. 2, an image processing method is provided, which is described by taking the method as an example applied to the server in fig. 1, and includes the following steps:
The external storage device may be a device for storing images other than the computing unit, and may be a hard disk, a mobile storage device, or the like.
The image to be processed can be an image which is stored in an external storage device and is not subjected to pixel shuffling, the image to be processed comprises at least one pixel unit, and the image to be processed is cut into a plurality of pixel units according to the principle of pixel point arrangement sequence.
The pixel unit can be one of the cut fragments of the image to be processed after cutting, and one pixel unit is composed of at least one pixel point.
The image pixel data can be data contained in each pixel point in the pixel unit, and for any pixel point, at least one channel correspondingly expresses the image pixel data corresponding to the pixel point.
Specifically, the pixel shuffling unit reads image pixel data corresponding to a pixel unit in an image to be processed from an external storage device through a communication network, and for any pixel unit, the pixel shuffling unit is composed of at least one pixel point, and for any pixel point, the pixel shuffling unit is composed of at least one channel which correspondingly expresses image pixel data corresponding to the pixel point.
For example, the pixel shuffling unit a reads out image pixel data corresponding to pixel units in an image to be processed from the external storage device B, and the read pixel units are quantized.
And 204, performing pixel shuffling on the image pixel data corresponding to the pixel unit to obtain a characteristic diagram corresponding to the pixel unit, and outputting the characteristic diagram corresponding to the pixel unit to an external storage device.
The pixel shuffling may be an action of processing a picture by sub-pixel convolution (sub-pixel convolution), which is a method for skillful images and feature maps upscale and is also a simplified form of normal convolution, and a strong assumption is added, so that a large amount of convolution operations are removed. The result of the sub-pixel convolution is generally a larger picture, which can be used in super-resolution calculation.
The feature map corresponding to the pixel unit may be an image that can represent the feature of the pixel unit and is obtained after the pixel unit is subjected to pixel shuffling, and the feature map corresponding to the general pixel unit is a larger picture and can represent information expressed by a plurality of channels corresponding to the pixel points in the pixel unit.
Specifically, each pixel unit is integrally input to a pixel shuffling unit, pixel shuffling is performed on image pixel data in the pixel unit, namely sub-pixel convolution is operated, a characteristic diagram corresponding to the pixel unit is obtained after the pixel shuffling, and the characteristic diagram corresponding to the pixel unit is output to an external storage device. In general, the feature map after the pixel shuffling is a larger image, and information included in each channel in the pixel shuffling can be expressed.
For example, the pixel units 1-10 are respectively input into a pixel shuffling unit a, the pixel shuffling unit a performs pixel shuffling on the input pixel units 1-10, that is, performs sub-pixel convolution to obtain feature maps corresponding to the pixel units 1-10, and outputs the feature maps corresponding to the pixel units 1-10 to an external storage device B.
In step 206, the pixel unit formed by the next pixel point of each pixel point in the pixel unit is used as the pixel unit.
The pixel point arrangement sequence can be a preset sequence, the sequence is used for cutting the image to be processed into pixel units and splicing the feature maps subjected to pixel shuffling, the sequence of the segmented pixel units is the address of the first pixel point in each pixel unit, and the sequence of the segmented pixel units is the address of the first image pixel data in each feature map for the feature maps subjected to pixel shuffling.
Specifically, after the previous pixel unit runs the pixel shuffling, the corresponding feature map is output to an external storage device, and meanwhile, the pixel shuffling unit reads the next pixel unit as the pixel unit to be processed based on the preset pixel point arrangement sequence.
For example, after the pixel shuffling is completed in the pixel shuffling unit No. 5, the corresponding feature map is output to the external storage device B, and the pixel shuffling unit a reads the pixel unit No. 6 as the pixel unit to be processed based on the preset pixel arrangement sequence.
And step 208, returning to the step of reading the image pixel data corresponding to the pixel units in the image to be processed from the external storage device until the feature maps corresponding to all the pixel units in the image to be processed are obtained.
Specifically, after the last pixel unit completes the operation of pixel shuffling, the next pixel unit is input to the pixel shuffling unit, the pixel shuffling unit repeatedly operates to input all the pixel units to the pixel shuffling unit, pixel shuffling is performed on image pixel data in the pixel units, namely sub-pixel convolution is operated, a characteristic image corresponding to the pixel unit is obtained after the pixels are shuffled, and the characteristic image corresponding to the pixel unit is output to an external storage device. In general, the feature map after pixel shuffling is a larger image, and information contained in each channel in the pixel shuffling can be expressed until all pixel units corresponding to the image to be processed are subjected to pixel shuffling to obtain a corresponding feature map. The characteristic graphs corresponding to all the pixel units are used for forming the pixel-shuffled image corresponding to the image to be processed.
For example, after the 5 th pixel unit completes the pixel shuffling operation, the 6 th pixel unit is input to the pixel shuffling unit, the pixel shuffling unit repeatedly inputs the whole pixel unit to the pixel shuffling unit, pixel shuffling is performed on image pixel data in the pixel unit, namely sub-pixel convolution is performed, a feature map corresponding to the pixel unit is obtained after the pixels are shuffled, and the feature map corresponding to the pixel unit is output to an external storage device. In general, the feature map after pixel shuffling is a larger image, and information contained in each channel in the pixel shuffling can be expressed until all 100 pixel units corresponding to the image to be processed are subjected to pixel shuffling to obtain the corresponding feature map. The characteristic map corresponding to 100 pixel units is used for forming a pixel-shuffled image corresponding to the image to be processed.
In the image processing method, image pixel data corresponding to pixel units in an image to be processed is read from an external storage device; wherein, the pixel unit is composed of at least one pixel point; performing pixel shuffling on image pixel data corresponding to the pixel units to obtain characteristic diagrams corresponding to the pixel units, and outputting the characteristic diagrams corresponding to the pixel units to an external storage device; pixel shuffling is the convolution processing of image pixel data; taking a pixel unit consisting of a next pixel point of each pixel point in the pixel units as a pixel unit; returning to the step of reading image pixel data corresponding to the pixel units in the image to be processed from the external storage device until feature maps corresponding to all the pixel units in the image to be processed are obtained; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed.
By adopting the method of arranging the pixel shuffling unit in the server, any pixel unit of the picture to be processed can be directly subjected to pixel shuffling, the newly proposed address operation is used, the pixel units which do not need to be processed can be temporarily stored in the external storage device, only the pixel units which are being processed need to be taken out of the external storage device, and the volatile storage resource of heterogeneous computing equipment is saved. Meanwhile, a large amount of shifting or similar operations are not needed to be carried out on data in volatile storage resources of the heterogeneous computing equipment, pixel shuffling can be completed directly at the stage of data taking out, and the data processing efficiency is greatly improved.
In some embodiments, as shown in fig. 3, when a pixel unit is composed of at least two pixels, performing pixel shuffling on image pixel data corresponding to the pixel unit to obtain a feature map corresponding to the pixel unit, including:
Specifically, the pixel shuffling unit reads image pixel data corresponding to pixel points in pixel units in an image to be processed from an external storage device through a communication network, and for any pixel unit, the pixel shuffling unit is composed of at least one pixel point, and for any pixel point, the pixel shuffling unit is composed of image pixel data corresponding to the pixel point, which is expressed by at least one channel. And integrally inputting each pixel unit into a pixel shuffling unit, performing pixel shuffling on image pixel data of pixel points in the pixel units, namely performing sub-pixel convolution, and obtaining a characteristic diagram corresponding to the pixel points after pixel shuffling. In general, the feature map after the pixel shuffling is a larger image, and information included in each channel in the pixel shuffling can be expressed.
For example, the pixel shuffling unit a reads out image pixel data corresponding to pixel points in pixel units in the image to be processed from the external storage device B, and the condition of the read pixel points in the pixel units has a quantization form. Respectively inputting pixel points 1-10 into a pixel shuffling unit A, and performing pixel shuffling on the input pixel points 1-10 by the pixel shuffling unit A, namely performing sub-pixel convolution to obtain a characteristic diagram corresponding to the pixel points 1-10
Specifically, after the current pixel is subjected to pixel shuffling, the corresponding characteristic graph is output to an external storage device, and meanwhile, the pixel shuffling unit reads the next pixel as the pixel to be processed based on the preset pixel arrangement sequence.
After the last pixel completes the operation of pixel shuffling, the next pixel is input to the pixel shuffling unit, the pixel shuffling unit repeatedly operates to input all the pixel units to the pixel shuffling unit, pixel shuffling is carried out on image pixel data of the pixels in the pixel units, namely sub-pixel convolution is operated, and a characteristic diagram corresponding to the pixel is obtained after pixel shuffling is carried out. Generally, the feature map after pixel shuffling is a larger image, and information contained in each channel in the pixel shuffling can be expressed until all pixel points corresponding to the image to be processed are subjected to pixel shuffling to obtain a corresponding feature map.
For example, after the pixel shuffling is completed by the 5 th pixel, the corresponding feature map is output to the external storage device B, and the pixel shuffling unit a reads the 6 th pixel as the pixel to be processed based on the preset pixel arrangement sequence.
After the 5 th pixel point finishes the operation of pixel shuffling, the 6 th pixel point is input to the pixel shuffling unit, the pixel shuffling unit repeatedly operates to input all the pixel units to the pixel shuffling unit, the pixel shuffling is carried out on the image pixel data of the pixel points in the pixel units, namely, sub-pixel convolution is operated, and the characteristic image corresponding to the pixel point is obtained after the pixel shuffling. Generally, the feature map after pixel shuffling is a larger image, and information contained in each channel in the pixel shuffling can be expressed until all 100 pixel points corresponding to the image to be processed are subjected to pixel shuffling to obtain a corresponding feature map.
Specifically, the feature maps corresponding to all pixel points are used to compose the pixel-shuffled image corresponding to the pixel unit.
For example, a feature map corresponding to 100 pixel points is used to compose a pixel-shuffled image corresponding to a pixel unit.
In this embodiment, all the pixel points in the pixel unit are subjected to pixel shuffling by the pixel shuffling unit, so that the feature maps corresponding to all the pixel points in the pixel unit can be obtained, and the definition of the sub-image to be processed corresponding to the pixel unit is improved.
In some embodiments, as shown in fig. 4, the image to be processed has a plurality of channels, and reading image pixel data corresponding to pixel units in the image to be processed from an external storage device includes:
Specifically, the pixel shuffling unit reads image pixel data corresponding to a pixel point in a current channel from an external storage device through a communication network, and for any pixel unit, the pixel shuffling unit is composed of at least one pixel point, and for any pixel point, the pixel shuffling unit is composed of image pixel data corresponding to the pixel point and expressed by at least one channel.
For example, the pixel shuffling unit a reads out image pixel data corresponding to a current channel of a pixel point from the external storage device B, and the read channel has a quantization form.
Specifically, after the previous channel finishes the operation of reading the image pixel data corresponding to the pixel point in the current channel, the next channel is input to the pixel shuffling unit, and the pixel shuffling unit repeatedly reads the image pixel data corresponding to the pixel point in the current channel from the external storage device through the communication network until the image pixel data corresponding to the pixel point in all the channels are read.
For example, after the 5 th channel finishes the operation of reading the image pixel data corresponding to the pixel point in the current channel, the 6 th channel is input to the pixel shuffling unit, and the pixel shuffling unit repeatedly reads the image pixel data corresponding to the pixel point in the current channel from the external storage device through the communication network until the image pixel data corresponding to the pixel point in all 100 channels are read.
And step 406, taking the image pixel data corresponding to the pixel points in all the channels as the image pixel data corresponding to the pixel points.
Specifically, the image pixel data of all channels corresponding to the pixel point are combined to obtain the image pixel data corresponding to the pixel point.
For example, the image pixel data of 100 channels corresponding to the pixel point are combined to obtain the image pixel data corresponding to the pixel point.
In this embodiment, the pixel shuffling unit extracts the image pixel data corresponding to all channels of one pixel point, so that the image pixel data corresponding to all channels in the pixel point can be obtained, and the definition of the sub-image to be processed corresponding to the pixel point is improved.
In some embodiments, as shown in fig. 5, before reading image pixel data corresponding to a pixel unit in an image to be processed from an external storage device, the method further includes:
The first pixel point may be a pixel point with the first address rank in each pixel point in each pixel unit.
The position relationship may be a relationship between an address of each pixel point in the pixel unit and the first pixel point.
Specifically, the pixel shuffling unit reads addresses of all pixel points in the pixel unit from the external storage device, and obtains the position relation between any one pixel point except the first pixel point and the first pixel point by taking the address of the first pixel point as a reference system.
For example, the pixel shuffling unit a reads addresses 1 to 100 corresponding to each (100) pixel points in the pixel unit 1 from the external storage device B, and obtains the positional relationship between the pixels 2 to 100 except the pixel 1 relative to the pixel 1by using the address of the pixel 1 as a reference system.
Specifically, according to the pixel address corresponding to the first pixel point and the position relationship between the addresses of other pixel points and the address of the first pixel point, the address corresponding to any one pixel point in the same pixel unit can be determined. The calculation formula of the position relation is as follows: addr =0 × 0a00 0000+20 (offset) × n (data _ width)/8 (1 byte), addr is a position relationship, and n is the number of pixels.
For example, according to the pixel address 1 corresponding to the first pixel point and the position relationship between the addresses 2-100 of other pixel points and the address 1 of the first pixel point, the addresses 2-100 corresponding to the pixel points 2-100 in the same pixel unit can be determined.
Specifically, after the addresses of the rest of the pixel points are further determined according to the address of the first pixel point in the pixel unit, the pixel shuffling unit reads image pixel data corresponding to the pixel unit in the image to be processed from the external storage device according to the determined pixel point address.
For example, after further determining addresses 2-100 of the remaining pixels according to the address 1 of the first pixel in the pixel unit, the pixel shuffling unit reads out image pixel data 1-100 corresponding to the pixel unit in the image to be processed from the external storage device according to the determined pixel address.
In this embodiment, by using the pixel address corresponding to the first pixel point and the position relationship between each pixel point and the first pixel point, the accurate address of any pixel point can be accurately located, and the efficiency of the pixel shuffling unit in reading the pixel points is improved.
In some embodiments, as shown in fig. 6, determining the pixel address corresponding to each pixel point according to the position relationship and the pixel address corresponding to the first pixel point includes:
Specifically, for any pixel unit in the image to be processed, any pixel point in a plurality of pixel points included in the pixel unit obtains a corresponding relationship between the pixel point and the first pixel point.
For example, for any pixel unit N in the image to be processed, which includes 100 pixels, for any pixel m, the corresponding relationship between the pixel m and the first pixel is obtained.
The address offset may be a distance between a selected arbitrary pixel and a reference pixel, and the distance is expressed by a difference between an address of the selected arbitrary pixel and an address of the reference pixel.
Specifically, according to the position relationship corresponding to any one pixel point in the pixel unit, the offset of any one pixel point relative to the address of the first pixel point is further determined, and the address of the first pixel point and the offset are added or subtracted to obtain the pixel address corresponding to any one pixel point in the pixel unit.
For example, according to the position relationship corresponding to any one pixel point in the pixel unit, the offset 2-100 of the pixel point 2-100 relative to the address of the first pixel point is further determined, and the pixel addresses 2-100 corresponding to the pixel points 2-100 in the pixel unit are obtained based on the address of the first pixel point and the addition or subtraction of the offsets 2-100 respectively.
In this embodiment, by introducing the address offset corresponding to the position relationship corresponding to any pixel point and the pixel address corresponding to the first pixel point, the address of any pixel point can be accurately obtained by adding the address offset and the pixel address corresponding to the first pixel point, and the reading accuracy of the address of any pixel point is improved.
In some embodiments, as shown in fig. 7, the method further comprises:
in step 702, a ratio between a full-load data processing amount corresponding to the pixel shuffling unit and a data amount corresponding to one pixel unit is obtained.
The full-load data processing amount may be a data processing amount corresponding to the pixel shuffling unit processing data by using all resources.
Specifically, the data processing amount that can be processed when the pixel shuffling unit is fully loaded and the data amount that is processed corresponding to one pixel unit are obtained, and the data processing amount that is fully loaded corresponding to the pixel shuffling unit is divided by the data amount that is corresponding to one pixel unit to obtain the ratio of the data processing amount and the data amount.
For example, if the full-load data throughput of the pixel shuffling unit is 10 and the data amount of one pixel unit is 2, the ratio of the full-load data throughput of the pixel shuffling unit to the data amount of one pixel unit is 5.
In step 704, if the ratio is greater than or equal to the threshold, the ratio is rounded as the number of multiple threads, so that the pixel shuffling unit executes the multi-thread calculation according to the number of multiple threads.
Specifically, if the ratio between the full-load data processing amount corresponding to the pixel shuffling unit and the data amount corresponding to one pixel unit is greater than or equal to a preset threshold value, the obtained ratio is rounded as the thread number of the multithread processing of the pixel shuffling unit, and the pixel shuffling unit is used for adopting multithread calculation according to the number of multiple threads.
For example, if the ratio between the full-load data processing amount corresponding to the pixel shuffling unit and the data amount corresponding to one pixel unit is 5.5 and is greater than the preset threshold value 2, the obtained ratio 5.5 is rounded, that is, 6 is used as the thread number 6 of the multithread processing of the pixel shuffling unit, and the pixel shuffling unit is used for adopting the multithread calculation according to the thread number 6.
In this embodiment, by using the ratio between the processing amount of the full-load data corresponding to the pixel shuffling unit and the data amount corresponding to one pixel unit, the number of the pixel shuffling units capable of processing the pixel units simultaneously can be known, and the pixel shuffling efficiency can be improved by further using multi-thread processing.
In some embodiments, DDR memory and addressing works as follows:
1. in DDR, data is stored in sequential order by an operator that has not undergone pixel shuffling. Generally, when data is read from a DDR, the data can be fetched by only giving the start address of the data and the required data length using the burst transfer mechanism (burst) of the AXI4 protocol. After the burst transfer obtains the start address and the length of the data to be transferred, the burst transfer starts from the start address and continuously takes out the data with the length required by us.
2. The data length required to be read is given according to the data quantity required each time, if only 9 data are required and the bit width of each data is 16-bit, the data length rd _ data _ length =9 (data _ number) × 16 (data _ width)/8 (1 byte) required to be read, and the output data length is in the unit of byte.
3. For the address of a single pixel, since the image is sequentially and continuously stored in the DDR, we need to know the address of the first pixel of the image first, and then calculate the address of the pixel according to the data bit width of the pixel and the offset of the pixel from the first pixel. For example, the address of the first pixel is base _ addr =0 × 0a00 0000, the data bit width of the single pixel is 16-bit, and the address of the 20 th pixel is obtained, so that the address of the 16 th pixel is addr =0 × 0a00 0000+20 (offset) × 16 (data _ width)/8 (1 byte). The calculation process is in fact simply summarized as data length of base address + offset, the base address being preset.
In some embodiments, pixel shuffling is accomplished by addressing. To complete the conversion of the pixel arrangement order, the address of the first pixel is given, the first pixel is read, the first pixel is written back to the new address of the DDR, the address of the second pixel is given, the second pixel is read, the DDR is written back, and the address written back to the DDR is adjacent to the new address. By analogy, the whole pixel shuffling process can be completed only by the control logic of address addressing without separately opening up the storage space of heterogeneous computing equipment to buffer the pictures. The mentioned extra memory space of heterogeneous computing device is mainly referred to as BRAM or SRAM, and the data taken each time is mainly buffered by the register/latch used in the control process.
In practice, only a single pixel is usually not transmitted, and thus the data transmission efficiency is greatly reduced. The amount of data actually transferred at a time depends on the model and the role that the operator plays in the model. In addition, when data is read according to the sequence or mode of pixel shuffling, the data does not need to be stored back to the DDR until complete pixel shuffling is finished, and the data can also be transmitted to a subsequent logic module for further processing.
It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially shown as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts according to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a part of the steps or stages in other steps.
Based on the same inventive concept, the embodiment of the application also provides an image processing device. The implementation scheme for solving the problem provided by the apparatus is similar to the implementation scheme described in the above method, so specific limitations in the following embodiments of the image processing apparatus may refer to the limitations on the image processing method in the foregoing, and details are not repeated herein.
In some embodiments, as shown in fig. 8, there is provided an image processing apparatus applied to a pixel shuffling unit connected to an external storage device, the external storage device storing images to be processed, comprising:
a reading module 802, configured to read image pixel data corresponding to a pixel unit in an image to be processed from an external storage device; wherein, the pixel unit is composed of at least one pixel point;
the shuffling module 804 is used for carrying out pixel shuffling on image pixel data corresponding to the pixel units to obtain a characteristic diagram corresponding to the pixel units and outputting the characteristic diagram corresponding to the pixel units to an external storage device; pixel shuffling is to perform convolution processing on image pixel data;
a replacing module 806, configured to use a pixel unit formed by a next pixel point of each pixel point in the pixel unit as a pixel unit;
a loop module 808, configured to return to the step of reading image pixel data corresponding to pixel units in the image to be processed from the external storage device until obtaining feature maps corresponding to all pixel units in the image to be processed; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed.
In some embodiments, when a pixel unit is composed of at least two pixels, pixel shuffling is performed on image pixel data corresponding to the pixel unit to obtain a feature map corresponding to the pixel unit, and the shuffling module is specifically configured to:
performing pixel shuffling on image pixel data corresponding to pixel points in the pixel units to obtain a feature map corresponding to the pixel points;
taking the next pixel point of the pixel points in the pixel units as a pixel point; returning to the step of performing pixel shuffling on image pixel data corresponding to the pixel points in the pixel unit to obtain a feature map corresponding to the pixel points until feature maps corresponding to all the pixel points in the pixel unit are obtained;
and determining the characteristic diagram corresponding to the pixel unit according to the characteristic diagrams corresponding to all the pixel points in the pixel unit.
In some embodiments, in the aspect that the image to be processed has a plurality of channels, and image pixel data corresponding to a pixel unit in the image to be processed is read from the external storage device, the reading module is specifically configured to:
for each pixel point in the pixel unit, reading image pixel data corresponding to the pixel point in the current channel from an external storage device;
taking the next channel of the current channel in the plurality of channels as the current channel; returning to the step of reading the image pixel data corresponding to the pixel points in the current channel from the external storage device until obtaining the image pixel data corresponding to the pixel points in all channels;
and taking the image pixel data corresponding to the pixel points in all the channels as the image pixel data corresponding to the pixel points.
In some embodiments, before reading image pixel data corresponding to a pixel unit in an image to be processed from the external storage device, the reading module is further configured to:
acquiring the position relation between each pixel point and a first pixel point in the pixel unit;
determining a pixel address corresponding to each pixel point according to the position relation and the pixel address corresponding to the first pixel point;
reading image pixel data corresponding to a pixel unit in an image to be processed from an external storage device, comprising:
and reading image pixel data corresponding to the pixel units in the image to be processed from the external storage equipment according to the pixel addresses corresponding to the pixel points.
In some embodiments, in determining the pixel address corresponding to each pixel point according to the position relationship and the pixel address corresponding to the first pixel point, the reading module is specifically configured to:
aiming at any pixel point in the pixel unit, acquiring a position relation corresponding to any pixel point;
and determining the pixel address corresponding to any pixel point according to the address offset corresponding to the position relation corresponding to any pixel point and the pixel address corresponding to the first pixel point.
In some embodiments, the shuffling module is further specifically configured to:
acquiring the ratio of the processing amount of full-load data corresponding to the pixel shuffling unit to the data amount corresponding to one pixel unit;
and if the ratio is larger than or equal to the threshold value, rounding the ratio to be used as the number of multiple threads, and enabling the pixel shuffling unit to execute the multi-thread calculation according to the number of the multiple threads.
The respective modules in the image processing apparatus described above may be wholly or partially implemented by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In some embodiments, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 9. The computer device includes a processor, a memory, an Input/Output interface (I/O for short), and a communication interface. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface is connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing server data. The communication interface of the computer device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement the steps of the image processing method described above.
Those skilled in the art will appreciate that the architecture shown in fig. 9 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In some embodiments, there is also provided a computer device comprising a memory in which a computer program is stored and a processor which, when executing the computer program, implements the steps of the image processing method described above.
In some embodiments, a computer-readable storage medium is provided, as shown in fig. 10, which stores a computer program that, when executed by a processor, implements the steps of the image processing method described above.
In some embodiments, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the image processing method described above.
It should be noted that, the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data need to comply with the relevant laws and regulations and standards of the relevant country and region.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above may be implemented by hardware instructions of a computer program, and the computer program may be stored in a non-volatile computer-readable storage medium, and when executed, may include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), magnetic Random Access Memory (MRAM), ferroelectric Random Access Memory (FRAM), phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), among others. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
The above examples only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, and these are all within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims (10)
1. An image processing method, applied to a pixel shuffling unit connected to an external storage device, the external storage device storing images to be processed, the method comprising:
reading image pixel data corresponding to pixel units in the image to be processed from the external storage equipment; the pixel unit consists of at least one pixel point;
pixel shuffling is carried out on image pixel data corresponding to the pixel unit to obtain a feature map corresponding to the pixel unit, and the feature map corresponding to the pixel unit is output to the external storage device; the pixel shuffling is convolution processing of the image pixel data;
taking a pixel unit composed of a next pixel point of each pixel point in the pixel units as the pixel unit;
returning to the step of reading the image pixel data corresponding to the pixel units in the image to be processed from the external storage device until obtaining the feature maps corresponding to all the pixel units in the image to be processed; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed.
2. The method of claim 1, wherein when the pixel unit is composed of at least two pixels, the pixel shuffling is performed on image pixel data corresponding to the pixel unit to obtain a feature map corresponding to the pixel unit, comprising:
performing pixel shuffling on image pixel data corresponding to pixel points in the pixel units to obtain a feature map corresponding to the pixel points;
taking a next pixel point of the pixel points in the pixel units as the pixel point; returning to the step of performing pixel shuffling on the image pixel data corresponding to the pixel points in the pixel unit to obtain the feature maps corresponding to the pixel points until the feature maps corresponding to all the pixel points in the pixel unit are obtained;
and determining the characteristic graph corresponding to the pixel unit according to the characteristic graphs corresponding to all the pixel points in the pixel unit.
3. The method according to claim 2, wherein the image to be processed has a plurality of channels, and the reading of the image pixel data corresponding to the pixel unit in the image to be processed from the external storage device comprises:
for each pixel point in the pixel unit, reading image pixel data corresponding to the pixel point in the current channel from the external storage device;
taking a next channel of the current channels of the plurality of channels as the current channel; returning to the step of reading the image pixel data corresponding to the pixel point in the current channel from the external storage device until obtaining the image pixel data corresponding to the pixel point in all the channels;
and taking the image pixel data corresponding to the pixel points in all the channels as the image pixel data corresponding to the pixel points.
4. The method according to claim 1, wherein before reading the image pixel data corresponding to the pixel unit in the image to be processed from the external storage device, the method further comprises:
acquiring the position relation between each pixel point and a first pixel point in the pixel unit;
determining a pixel address corresponding to each pixel point according to the position relation and the pixel address corresponding to the first pixel point;
the reading of the image pixel data corresponding to the pixel unit in the image to be processed from the external storage device includes:
and reading image pixel data corresponding to the pixel units in the image to be processed from the external storage equipment according to the pixel addresses corresponding to the pixel points.
5. The method according to claim 4, wherein the determining the pixel address corresponding to each pixel point according to the position relationship and the pixel address corresponding to the first pixel point comprises:
aiming at any pixel point in the pixel unit, acquiring a position relation corresponding to the pixel point;
and determining the pixel address corresponding to any pixel point according to the address offset corresponding to the position relation corresponding to any pixel point and the pixel address corresponding to the first pixel point.
6. The method according to any one of claims 1 to 5, further comprising:
acquiring the ratio between the processing amount of full-load data corresponding to the pixel shuffling unit and the data amount corresponding to one pixel unit;
and if the ratio is larger than or equal to the threshold value, rounding the ratio to be the number of multiple threads, and enabling the pixel shuffling unit to execute multi-thread calculation according to the number of the multiple threads.
7. An image processing apparatus applied to a pixel shuffling unit connected to an external storage device storing an image to be processed, comprising:
the reading module is used for reading image pixel data corresponding to the pixel units in the image to be processed from the external storage equipment; the pixel unit is composed of at least one pixel point;
the shuffling module is used for carrying out pixel shuffling on image pixel data corresponding to the pixel unit to obtain a characteristic diagram corresponding to the pixel unit and outputting the characteristic diagram corresponding to the pixel unit to the external storage device; the pixel shuffling is convolution processing of the image pixel data;
the replacing module is used for taking a pixel unit formed by the next pixel point of each pixel point in the pixel unit as the pixel unit;
the circulation module is used for returning to the step of reading the image pixel data corresponding to the pixel units in the image to be processed from the external storage device until obtaining the feature maps corresponding to all the pixel units in the image to be processed; and the characteristic diagrams corresponding to all the pixel units in the image to be processed are used for forming the image after pixel shuffling corresponding to the image to be processed.
8. The apparatus of claim 7, wherein in the aspect that when the pixel unit is composed of at least two pixels, the pixel shuffling module is specifically configured to shuffle the image pixel data corresponding to the pixel unit to obtain the feature map corresponding to the pixel unit:
performing pixel shuffling on image pixel data corresponding to pixel points in the pixel units to obtain a feature map corresponding to the pixel points;
taking the next pixel point of the pixel points in the pixel units as the pixel point; returning to the step of performing pixel shuffling on the image pixel data corresponding to the pixel points in the pixel unit to obtain the feature maps corresponding to the pixel points until the feature maps corresponding to all the pixel points in the pixel unit are obtained;
and determining the characteristic graph corresponding to the pixel unit according to the characteristic graphs corresponding to all the pixel points in the pixel unit.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 6.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211127995.9A CN115456858B (en) | 2022-09-16 | 2022-09-16 | Image processing method, device, computer equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211127995.9A CN115456858B (en) | 2022-09-16 | 2022-09-16 | Image processing method, device, computer equipment and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115456858A true CN115456858A (en) | 2022-12-09 |
| CN115456858B CN115456858B (en) | 2023-07-18 |
Family
ID=84304727
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211127995.9A Active CN115456858B (en) | 2022-09-16 | 2022-09-16 | Image processing method, device, computer equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115456858B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106328045A (en) * | 2015-07-06 | 2017-01-11 | 西安诺瓦电子科技有限公司 | Programmable logic device, sampling method under sub pixel, and related application |
| CN111897579A (en) * | 2020-08-18 | 2020-11-06 | 腾讯科技（深圳）有限公司 | Image data processing method, image data processing device, computer equipment and storage medium |
| CN111915673A (en) * | 2020-07-22 | 2020-11-10 | 深圳云天励飞技术有限公司 | Image processing method, image processing device, terminal equipment and storage medium |
| CN112749709A (en) * | 2019-10-30 | 2021-05-04 | 浙江商汤科技开发有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN114758209A (en) * | 2022-06-14 | 2022-07-15 | 深圳思谋信息科技有限公司 | Convolution result obtaining method and device, computer equipment and storage medium |
-
2022
- 2022-09-16 CN CN202211127995.9A patent/CN115456858B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106328045A (en) * | 2015-07-06 | 2017-01-11 | 西安诺瓦电子科技有限公司 | Programmable logic device, sampling method under sub pixel, and related application |
| CN112749709A (en) * | 2019-10-30 | 2021-05-04 | 浙江商汤科技开发有限公司 | Image processing method and device, electronic equipment and storage medium |
| CN111915673A (en) * | 2020-07-22 | 2020-11-10 | 深圳云天励飞技术有限公司 | Image processing method, image processing device, terminal equipment and storage medium |
| CN111897579A (en) * | 2020-08-18 | 2020-11-06 | 腾讯科技（深圳）有限公司 | Image data processing method, image data processing device, computer equipment and storage medium |
| CN114758209A (en) * | 2022-06-14 | 2022-07-15 | 深圳思谋信息科技有限公司 | Convolution result obtaining method and device, computer equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115456858B (en) | 2023-07-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109194960B (en) | Image frame rendering method and device and electronic equipment | |
| CN110321958B (en) | Training method of neural network model and video similarity determination method | |
| CN112001914A (en) | Depth image completion method and device | |
| CN109857744B (en) | Sparse tensor calculation method, device, equipment and storage medium | |
| KR20210036715A (en) | Neural processing apparatus and method for processing pooling of neural network thereof | |
| CN114266945B (en) | Training method of target detection model, target detection method and related device | |
| US20200327638A1 (en) | Connected component detection method, circuit, device and computer-readable storage medium | |
| CN115082358A (en) | Image enhancement method and device, computer equipment and storage medium | |
| CN115470156A (en) | RDMA-based memory use method, system, electronic device and storage medium | |
| US10049425B2 (en) | Merging filters for a graphic processing unit | |
| WO2024140568A1 (en) | Image processing method and apparatus, electronic device, and readable storage medium | |
| CN117370488A (en) | Data processing method, device, electronic equipment and computer readable storage medium | |
| CN116051345A (en) | Image data processing method, device, computer equipment and readable storage medium | |
| CN115456858B (en) | Image processing method, device, computer equipment and computer readable storage medium | |
| CN114140618A (en) | Convolution characteristic caching method and device, electronic equipment and readable storage medium | |
| CN114897702A (en) | Image reconstruction method, image reconstruction device, computer equipment and storage medium | |
| CN109375952B (en) | Method and apparatus for storing data | |
| CN115272706B (en) | Image processing method, device, computer equipment and storage medium | |
| CN113111891B (en) | Image reconstruction method and device, terminal equipment and storage medium | |
| CN116659520B (en) | Matching positioning method, device and equipment based on bionic polarization vision enhancement | |
| CN115665335B (en) | Image processing method, image processing apparatus, image forming apparatus, and medium | |
| US20230070730A1 (en) | Word based channels last ordering in memory | |
| CN116527908A (en) | Motion field estimation method, motion field estimation device, computer device and storage medium | |
| CN115908115A (en) | Face image processing method, live image processing method and device and electronic equipment | |
| CN117729296A (en) | Video generation method, apparatus, computer device, storage medium, and program product |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |