Image processing circuit and image enhancement method performed by the device

Abstract

The present invention provides an image processing circuit and an image enhancement method performed by a device, wherein an image processing circuit includes: an attribute recognition engine for recognizing attributes from an input image based on a model stored in a memory; a picture quality (PQ) engine for enhancing the input image based on the attribute recognized by the attribute recognition engine to generate an output image for display; a data collection module for generating a labeled image based on the attribute recognized by the attribute recognition engine and at least one of the input image or the output image, and adding the labeled image to a training database stored in the memory; and a training engine for retraining the model in the memory using the training database. The implementation of the present invention can provide greater flexibility in image quality adjustment.

Description

Technical Field

[0001] Embodiments of the present invention relate to image processing technology, and more particularly to a device and method having image processing capabilities for enhancing image quality. Background Art

[0002] Modern devices with image display capabilities often perform image enhancement operations when displaying images. For example, a television can enhance images or videos to be displayed on the screen, and a smartphone can enhance images or videos captured by or displayed on a smartphone. However, conventional devices often perform image enhancement operations based on algorithms or formulas preconfigured by the device manufacturer. There is limited flexibility in adjusting the algorithms or formulas after the consumer uses the device. Therefore, there is a need to improve the design of image processing devices to allow for greater flexibility in image quality adjustments. Summary of the invention

[0003] The invention provides an image processing circuit and an image enhancement method executed by the device, which has greater flexibility in image quality adjustment.

[0004] An image processing circuit provided by the present invention includes: an attribute recognition engine, which is used to recognize attributes from an input image based on a model stored in a memory; a picture quality (PQ) engine, which is used to enhance the input image based on the attributes recognized by the attribute recognition engine to generate an output image for display; a data collection module, which is used to generate a labeled image based on the attributes recognized by the attribute recognition engine and at least one of the input image or the output image, and add the labeled image to a training database stored in the memory; and a training engine, which uses the training database to retrain the model in the memory.

[0005] The present invention provides an image enhancement method performed by a device, comprising: identifying attributes from an input image based on a model stored in the device; generating an output image for display by enhancing the input image based on the identified attributes; generating a labeled image based on the attributes identified by the model in the device and at least one of the input image or the output image; adding the labeled image to a training database stored in the device; and retraining the model stored in the device using the training database.

[0006] Another image enhancement method provided by the present invention and performed by a device includes: receiving attributes of an input image recognized by a user through a user interface; enhancing the input image based on the recognized attributes to generate an output image for display; generating a labeled image based on the attributes recognized by the user and at least one of the input image or the output image; adding the labeled image to a training database; retraining a model stored in the device using the training database; and enhancing the input image through the retrained model to generate an output image for display.

[0007] As described above, in the present application, a labeled image is generated through the identified attributes and at least one of the input image or the output image, and the generated labeled image is added to a training database, and then the model for identifying the attributes of the input image is retrained according to the updated training database, so that the model for identifying the attributes of the input image can be updated, thereby having greater flexibility in image quality adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 is a block diagram illustrating an image processing circuit that performs image quality enhancement according to one embodiment.

[0009] Figure 2 is a block diagram illustrating automatic tagging according to one embodiment.

[0010] Figure 3 is a block diagram illustrating manual labeling according to one embodiment.

[0011] Figure 4A An example of an image with incorrectly identified scene type attributes is shown in accordance with one embodiment.

[0012] Figure 4B An example of a sample image for manual labeling according to one embodiment is shown.

[0013] Figure 5 is a flow chart illustrating a method for a device to perform an image enhancement operation according to one embodiment.

[0014] Figure 6 An example of a device according to one embodiment is shown. DETAILED DESCRIPTION

[0015] Certain words are used in the specification and claims to refer to specific components. It should be understood by those skilled in the art that hardware manufacturers may use different terms to refer to the same component. The specification and claims do not use differences in names as a way to distinguish components, but use differences in the functions of components as the criteria for distinction. "Including" and "including" mentioned throughout the specification and claims are open-ended terms, so they should be interpreted as "including but not limited to". "Substantially" means that within an acceptable error range, those skilled in the art can solve the technical problem within a certain error range and basically achieve the technical effect. In addition, the term "coupled" includes any direct and indirect electrical connection means. Therefore, if the text describes a first device coupled to a second device, it means that the first device can be directly electrically connected to the second device, or indirectly electrically connected to the second device through other devices or connection means. The following is a preferred way to implement the present invention, the purpose of which is to illustrate the spirit of the present invention rather than to limit the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the definition of the claims.

[0016] The following description is the best embodiment of the present invention. These descriptions are used to illustrate the general principles of the present invention and should not be used to limit the present invention. The scope of protection of the present invention should be determined based on the claims of the present invention.

[0017] This article describes a device including an image processing circuit. A user can view an image (e.g., a video) on a display panel coupled to the image processing circuit. The image processing circuit generates a training database containing automatically and / or manually labeled (labeled) images. The image processing circuit further uses the training database to retrain one or more models to identify one or more attributes of the image based on the model. The picture quality (Picture Quality, PQ) engine improves the quality of the output image by changing certain image values ​​associated with the identified (identified) attributes. If the user is not satisfied with the quality of the output image displayed on the display panel, the user can provide feedback to the device to help retrain the model used to generate the identified attributes. Therefore, users can customize the training database and model according to their viewing experience and preferences, and therefore, the device provides flexibility in image quality adjustment.

[0018] Figure 11 is a block diagram illustrating an image processing circuit 100 that performs image quality enhancement according to one embodiment. The image processing circuit 100 may be part of a device, also referred to as an edge device, such as a television, a smart phone, a computing device, a device connected to a network, a gaming device, an entertainment device, an Internet-of-things (IoT) device, or any device capable of processing and displaying images and / or videos. The images to be processed may be captured by the same device or by different sources and then downloaded, streamed, transmitted, or otherwise made accessible to the device. In some embodiments, the image processing circuit 100 may include an artificial intelligence (AI) processor 180 coupled to a picture quality (PQ) engine 140. The AI ​​processor 180 may be trained to infer representative characteristics in an input image, and the PQ engine 140 may improve the image quality of the input image based on the representative characteristics.

[0019] The image processing circuit 100 includes an input port 135 for receiving an input image 131 and an output port 145 for providing an output image 141, in this example, the output image 141 is an image processed by the input image 131. The output image 141 is sent to the display panel 160 for display. For ease of description, an input image and its corresponding output image are provided as examples. It should be understood that when the image processing circuit 100 receives an image sequence (e.g., a video) as input and generates a corresponding image sequence as output, the following description is also applicable.

[0020] The image processing circuit 100 also includes a control module 110 that sends control signals (shown in dotted lines) to control and manage training and inference operations on the device. The control module 110 triggers a training operation performed by the training engine 120 to train or retrain the model 125 using labeled images from the training database 155. The control module 110 also triggers an inference operation performed by the attribute recognition engine 130 to identify the attributes (i.e., representative features) of the input image 131. In one embodiment, the attribute recognition engine 130 can identify attributes by inference and / or measurement based on one or more models 125. The attribute identified by the attribute recognition engine 130 can be a type (e.g., scene type or object type), statistical information, or a feature in the image content. For example, the attribute can include a scene type, an object type in the scene, contrast information (e.g., a histogram or statistical data), brightness information (e.g., a histogram or statistical data), edge direction or strength, noise and blur, segmentation information, motion information, etc. In some embodiments, machine learning or deep learning algorithms can be used to identify attributes.

[0021] In some embodiments, the image processing circuit 100 may be implemented in a system on chip (SoC). In some embodiments, the image processing circuit 100 may be implemented in more than one chip in the same electronic device.

[0022] In one embodiment, the attribute recognition engine 130 can recognize multiple attributes (e.g., scene type and contrast information) from the image, which are collectively referred to as an attribute set of the image. The attribute recognition engine 130 can further generate a confidence level for the identified attribute; for example, the confidence level for the natural scene type is 75%. A high confidence level (e.g., when the confidence level exceeds a threshold) indicates that the attribute has a correspondingly high probability of being correctly recognized. The attribute recognition engine 130 sends the attribute set to the PQ engine 140 and the data collection module 150.

[0023] In one embodiment, the PQ engine 140 performs image enhancement operations on the input image 131 using an image processing algorithm based on a set of attributes of the input image 131. Different algorithms may be used for different attributes. For example, one algorithm is used for noise reduction, another algorithm is used for natural scenes, and another algorithm is used for scene types of food. In some embodiments, the PQ engine 140 may perform one or more of the following operations: noise reduction, scaling, contrast adjustment, color adjustment, and clarity adjustment. For example, the PQ engine 140 may increase the warmth of image colors in food scenes, increase clarity in blurred images, and denoise noisy images. The output of the PQ engine 140 is an output image 141, which is sent to the data collection module 150 and the output port 145.

[0024] The data collection module 150 receives the output image 141 from the PQ engine 140, and also receives the input image 131 and the attribute set of the input image 131 from the attribute recognition engine 130. In one embodiment, one or more recognized attributes in the attribute set may be accompanied by respective confidences. In other embodiments, only one of the input image 131 and the output image 141 is sent to the data collection module 150.

[0025] The data collection module 150 is a part of the image processing circuit 100, and the data collection module 150 provides the labeled images to the training database 155. In the manual labeling method, the input image 131 or the output image 141 is labeled by the user. In the automatic labeling method, the input image 131 or the output image 141 is automatically labeled with the attributes identified with high confidence. The automatic labeling and manual labeling will be referred to as Figure 2 and Figure 3 Give a description.

[0026] The control module 110 can trigger the training engine 120 to perform a training operation to train and / or retrain the model 125 using labeled images from the training database 155. The training operation can be performed periodically or based on an event. For example, the training operation can start when the image processing circuit 100 enters a sleep state or an idle state. For edge devices with limited processing resources (e.g., smart TVs, smart phones, IoT devices, etc.), the model 125 can be initially trained on a server such as a cloud server and then retrained by the training engine 120 on the edge device based on the viewed images or videos. The training operation changes the weights or parameters in the model 125, such as filter weights in an image filter, kernel weights in a neural network kernel, thresholds, etc. In some embodiments, the training operation can be performed by machine learning, deep learning, or other types of learning operations.

[0027] Figure 2is a block diagram illustrating automatic tagging according to one embodiment. Figure 2 , only the inference part of the image processing circuit 100 is shown. In this embodiment, the data collection module 150 includes a confidence check circuit 210, which compares the confidence of the identified attribute of the input image with a threshold. Different attributes can be compared with different thresholds. The data collection module 150 marks at least one of the input image or the output image with the identified attribute having a confidence exceeding the confidence threshold to generate a marked image, and uses the marked image to update the training database 155 for on-device training operations. For example, the input image can be identified as having the following attributes: urban street scene type, object types of cars, people and buildings, and a contrast of 80%. If the identified urban street scene type has a high confidence (i.e., exceeds its confidence threshold), the data collection module 150 will mark the input image or the output image as having an urban street scene type. The data collection module 150 can further mark the input image or the output image with other attributes having a confidence exceeding their respective confidence thresholds. The marked image is then stored in the training database 155.

[0028] Figure 3 is a block diagram illustrating manual labeling according to one embodiment. Figure 3 , only the inference portion of the image processing circuit 100 is shown. In this embodiment, the data collection module 150 is coupled to the user interface 320 to receive user feedback on the displayed image. The user interface 320 can provide a graphical user interface (GUI) and / or a voice control user interface on the display panel for receiving user voice commands, or provide other devices for receiving user input related to the displayed image. Via the user interface 320, the user can indicate that the image quality of one or more images is poor and request to start the manual marking process. Also refer to Figure 1 , the output image 141 is an example of a display image. In the following description, the terms "display image" and "output image" are used interchangeably.

[0029] In one embodiment, manual labeling can be performed based on a user's command. The user may label the displayed image as having poor image quality; for example, by selecting a button, the labeling operation will trigger the start of the manual labeling process. Alternatively, the user can request a manual labeling process for any image attribute at any time. In response, the image processing circuit 100 requests the user to label the displayed image (i.e., the output image) or the corresponding input image with the correct value or attribute type, where the "correctness" can be determined from the user's perspective. In one embodiment, the user interface 320 can present many optional values ​​or types to the user to replace the attributes recognized by the device. Taking the scene type as an example, the user interface 320 can present options such as "people", "food", "nature", "landmark" to the user to select as the scene type attribute of the image. The user can select one of the presented types (e.g., people) to indicate the correct scene type attribute of the image. In one embodiment, the user can add a new attribute such as "animal" to indicate the correct scene type attribute of the image.

[0030] In order to improve the training accuracy, the data collection module 150 can obtain multiple sample images from the training database 155, which are similar to the images marked by the user in terms of the attributes of interest. In one embodiment, the data collection module 150 includes a sample selection circuit 330, which selects sample images from the training database 155 and provides the selected sample images to the user. In terms of the attributes of interest, each selected sample image has a confidence level exceeding a predetermined threshold. For example, the sample image can be displayed on a display panel together with a list of optional values ​​or types of the attributes of interest. The user can mark the sample image by selecting a value or type from the list. In some embodiments, the user can add a new value or a new type to the attribute of interest. Using the above example where the scene type is an attribute of interest, each sample image can present a list of "people", "food", "nature", "landmark" for the user to select. The user can select from the list. Alternatively, the user can add "animals" to the list as a new option for the scene type. When the user has marked all the sample images provided by the sample selection circuit 330, the manual marking process ends.

[0031] Figure 4A 4 shows an example of an output image with an erroneously identified scene type attribute according to one embodiment. The display panel 400 may be coupled to the image processing circuit 100 ( Figure 1) output port 145. A user can mark the person image 410 as having poor image quality; for example, by selecting a button or by voice or other means when the image 410 is displayed. In response, the image processing circuit 100 presents a list of scene types for the scene type attribute when presenting the image 410. The image processing circuit 100 can also present values, types, or information of other attributes of the image 410 to the user in a previous or subsequent list. For the scene type attribute, the display panel 400 displays an automatically generated confidence for each scene type when displaying the image 410, for example, a person (0.05), a vehicle (0.62), a tree (0.77), a drink (0.02), and a house (0.23), where each number in brackets represents the confidence of the corresponding scene type in the image. Based on the confidence, the attribute recognition engine 130 identifies the tree scene type. In this example, the tree scene type is the identified attribute. Based on the identified attribute, the PQ engine 140 can apply an algorithm for enhancing the tree scene to generate an output image.

[0032] However, the user can determine from image 410 that the correct scene type attribute should be "people". For this image 410, the people scene type is a user-recognized attribute that is different from the tree scene type attribute recognized by the device. In this example, the user can select the "people" attribute to change the attribute recognized by the device for image 410. The user-recognized people scene type becomes the label of image 410.

[0033] After the user marks the image 410 with the corrected attribute, the image processing circuit 100 can present a plurality of sample images previously identified as human scene types to the user. The user can mark the scene types of these sample images to indicate whether they are correctly identified as containing human scene types. Figure 4B A sample image 420 is shown as an example. The data collection module 140 stores the user-labeled images including the user-labeled image 410 and the user-labeled sample image 420 into the training database 155 .

[0034] The training engine 120 uses the labeled images from the training database 155 to retrain the model 125. The training model 125 can be trained to detect features in an image or image sequence (e.g., edge direction and strength, segmentation information, motion, etc.), classify image content, measure the state of the image (e.g., contrast, sharpness, brightness, grayscale, noise, etc.), etc. The model 125 can be described by a mathematical formula or representation. The model 125 can be initially installed in the image processing circuit 100 and retrained or refined with labeled images to learn from the user's image viewing experience on the device.

[0035] Figure 5is a flow chart illustrating a method 500 for image enhancement according to one embodiment. In some embodiments, the image enhancement operations described herein include generating training data and retraining a model. The method 500 may be, for example, Figure 1 The image processing circuit 100 and / or Figure 6 It should be understood that Figure 1 and Figure 6 This is for illustrative purposes only. Other image processing devices may also perform the method 500 .

[0036] Method 500 begins at step 510. In step 510, the device identifies attributes from an input image based on a model stored in the device. In step 520, the device enhances the input image based on the identified attributes to generate an output image for display. In step 530, the device generates a labeled image based on the identified attributes and at least one of the input image or the output image (for example, the device generates a labeled input image based on the identified attributes and the input image, or the device generates a labeled output image based on the identified attributes and the output image). In step 540, the device adds the labeled image to a training database stored in the device. In step 550, the device retrains the model using the training database. In one embodiment, the model can be retrained on the device. In a specific implementation, Figure 5 The process is a loop process, that is, after step 550 is executed, the method of the present invention returns to step 510, and step 510 further recognizes attributes from the input image based on the model retrained in step 550. The loop process of method 500 can be stopped according to a preset condition (for example, when the attributes recognized for a predetermined number of times do not change) or stopped according to the user's instruction.

[0037] Figure 6 An example of a device 600 according to one embodiment is shown. The device 600 may include Figure 1 The image processing circuit 100 performs the above-mentioned image enhancement operation. The device 600 includes processing hardware 610. In one embodiment, the processing hardware 610 may include one or more processors, such as a central processing unit (CPU), a graphics processing unit (GPU), a digital processing unit (DSP), a multimedia processor, and other general and / or special processing circuits. In one embodiment, the processing hardware 610 may include hardware circuits, including but not limited to: Figure 1 The hardware 610 may include an artificial intelligence (AI) processor 615, which may be Figure 1 An example of an AI processor 180 in FIG. Return to reference Figure 1In one embodiment, the training engine 120 and / or the attribute recognition engine 130 may be part of the AI ​​processor 615. In one embodiment, the AI ​​processor 615 may be part of a GPU. In one embodiment, the AI ​​processor 615 may include a hardware accelerator, such as a convolutional neural network (CNN) accelerator 612.

[0038] CNN accelerator 612 includes hardware components specifically for accelerating neural network operations through convolution operations, fully connected operations, activation, pooling, normalization, element-by-element mathematical calculations, etc. In some embodiments, CNN accelerator 612 includes multiple computing units and memories (e.g., static random access memory (SRAM), wherein each computing unit further includes multiplier and adder circuits, etc., for performing mathematical operations, such as multiplication and accumulation (MAC) operations, to accelerate convolution, activation, pooling, normalization, and other neural network operations. CNN accelerator 612 can perform fixed-point and floating-point neural network operations. In conjunction with the image quality enhancement described herein, CNN accelerator 612 can perform combined Figure 1 Describes the training and inference operations.

[0039] Device 600 also includes memory and storage hardware 620 coupled to processing hardware 610. Memory and storage hardware 620 may include memory devices such as dynamic random access memory (DRAM), SRAM, flash memory, and other non-transitory machine-readable storage media (e.g., volatile or non-volatile storage devices). Memory and storage hardware 620 may further include storage devices, such as any type of solid-state or magnetic storage device. In one embodiment, memory and storage hardware 620 may store Figure 1 Model 125 and training database 155. In some embodiments, memory and storage hardware 620 may store instructions that, when executed by processing hardware 610, cause processing hardware 610 to perform the above-described image enhancement operations (e.g., Figure 5 Method 500).

[0040] The device 600 may also include a display panel 630 to display information such as images, videos, messages, web pages, games, text, and other types. Images may be marked by a user via a user interface such as a keyboard, a touchpad, a touch screen, a mouse, etc. The device 600 may also include audio hardware 640, such as a microphone and a speaker, for receiving and generating sound. The audio hardware 640 may also provide a user interface for sending and receiving voice commands.

[0041] In some embodiments, the device 600 may also include a network interface 650 to connect to a wired and / or wireless network for sending and / or receiving voice, digital data and / or media signals. It will be appreciated that for illustration purposes, the present invention is simplified. Figure 6 Embodiment of the invention. Figure 6 Embodiments of the invention may include other hardware components.

[0042] Already referenced Figure 1 and Figure 6 The exemplary embodiments describe Figure 5 However, it should be understood that Figure 5 The operations of the flowchart can be performed by Figure 1 and Figure 6 Embodiments other than the embodiments of the present invention are performed, and Figure 1 and Figure 6 Embodiments of the invention may perform operations different from those discussed with reference to the flowcharts. Figure 5 The flowcharts illustrate a particular order of operations performed by certain embodiments of the present invention, but it should be understood that such order is exemplary (e.g., alternative embodiments may perform operations in a different order, combine certain operations, overlap certain operations, etc.).

[0043] Although the present invention is disclosed as above with preferred embodiments, it is not intended to limit the scope of the present invention. Any person skilled in the art may make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be determined by the claims.

Claims

1. An image processing circuit, characterized in that: include: an attribute recognition engine for recognizing attributes from an input image based on a model stored in a memory; an image quality engine for enhancing the input image based on the attributes identified by the attribute recognition engine to generate an output image for display; a data collection module, configured to generate a labeled image based on the attribute identified by the attribute recognition engine and at least one of the input image or the output image, and add the labeled image to a training database stored in the memory; and A training engine, using the training database to retrain the model in the memory; The data collection module is also used for: When the confidence of the attribute identified by the attribute recognition engine exceeds a predetermined threshold, automatically generating the marked image using the attribute identified by the attribute recognition engine and at least one of the input image or the output image; and The training database is updated using the automatically labeled images.

2. The image processing circuit according to claim 1, characterized in that: Also included is an artificial intelligence processor, which at least partially includes the attribute recognition engine, and the attribute recognition engine is operable to execute a machine learning or deep learning algorithm to recognize the attribute.

3. The image processing circuit according to claim 1, wherein: A control module is also included to control the retraining of the model on a device including the image processing circuit based on an event or a periodic schedule.

4. The image processing circuit according to claim 1, wherein: The data collection module is also used for: receiving, through a user interface, an attribute recognized by a user from the input image, wherein the attribute recognized by the user changes the attribute recognized by the attribute recognition engine; and The marked image is generated based on the attribute identified by the user and at least one of the input image or the output image.

5. The image processing circuit according to claim 4, characterized in that: The user interface provides a list of options regarding attributes for selection by the user in response to an instruction from the user.

6. The image processing circuit according to claim 4, characterized in that: The data collection module is also used for: Acquire a plurality of sample images from the training database, each of the sample images having a confidence level exceeding a predetermined threshold in relation to the attribute identified by the user; and Each of the sample images is provided for user labeling, thereby generating the labeled images for the training database.

7. The image processing circuit according to claim 1, wherein: The attributes of the input image include one or more of the following: scene type, object type in the scene, contrast information, brightness information, edge direction or strength, noise information, blur level, segmentation information, and motion information.

8. The image processing circuit according to claim 1, wherein: The attribute recognition engine is further configured to recognize a plurality of attributes from the image sequence based on a plurality of models, wherein each model is configured to recognize one of the attributes.

9. The image processing circuit according to claim 1, wherein: The image quality engine is operable to perform image processing including one or more of the following: Denoising, scaling, contrast adjustment, color adjustment, and sharpness adjustment.

10. An image enhancement method performed by a device, characterized in that: include: recognizing attributes from an input image based on a model stored in the device; generating an output image for display by enhancing the input image based on the identified attributes; generating a labeled image based on the attributes identified by the model in the device and at least one of the input image or the output image; adding the labeled image to a training database stored in the device; and retraining the model stored in the device using the training database; The generating of the marked image further comprises: When the confidence of the attribute identified based on the model in the device exceeds a predetermined threshold, automatically generating the labeled image using the identified attribute and at least one of the input image or the output image; and The training database is updated using the automatically labeled images.

11. The image enhancement method performed by a device according to claim 10, characterized in that: The identification attributes also include: The attributes are identified using a machine learning or deep learning algorithm.

12. The image enhancement method performed by a device according to claim 10, characterized in that: Also includes: The model is retrained on the device based on an event or periodic schedule.

13. The image enhancement method performed by a device according to claim 10, characterized in that: The generating of the marked image further comprises: receiving an attribute identified by a user from the input image, the attribute identified by the user changing the attribute identified based on the model in the device; and The marked image is generated based on the attribute identified by the user and at least one of the input image or the output image.

14. The image enhancement method performed by a device according to claim 13, characterized in that: Also includes: In response to an instruction from the user, a list of options regarding attributes is displayed for selection by the user.

15. The image enhancement method performed by a device according to claim 13, characterized in that: Also includes: Acquire a plurality of sample images from the training database, each of the sample images having a confidence level exceeding a predetermined threshold in relation to the attribute identified by the user; and Each of the sample images is displayed for labeling by a user, thereby generating the labeled images for the training database.

16. The image enhancement method performed by a device as claimed in claim 10, characterized in that: The attributes of the input image include one or more of the following: scene type, object type in the scene, contrast information, brightness information, edge direction or strength, noise information, blur level, segmentation information, and motion information.

17. The image enhancement method performed by a device according to claim 10, characterized in that: Also includes: A plurality of attributes are identified from the image sequence based on a plurality of models, wherein each model is used to identify one of the attributes.

18. The image enhancement method performed by a device as claimed in claim 10, characterized in that: Also includes: Processing the input image, wherein processing the input image includes performing one or more of: denoising, scaling, contrast adjustment, color adjustment, and sharpness adjustment.

19. An image enhancement method performed by a device, characterized in that: include: receiving, via a user interface, attributes of an input image identified by a user; generating an output image for display by enhancing the input image based on the identified attributes; generating a labeled image based on the attribute identified by the user and at least one of the input image or the output image; adding the labeled images to a training database; retraining the model stored in the device using the training database; and enhancing the input image by the retrained model to generate an output image for display; When the confidence level of the attribute identified by the user exceeds a predetermined threshold, automatically generating the marked image using the attribute identified by the user and at least one of the input image or the output image; and The training database is updated using the automatically labeled images.

20. The image enhancement method performed by a device as claimed in claim 19, characterized in that: Also includes: In response to an instruction from the user, a list of options regarding attributes is displayed for selection by the user.

21. The image enhancement method performed by a device as claimed in claim 19, characterized in that: Also includes: Acquire a plurality of sample images from the training database, each of the sample images having a confidence level exceeding a predetermined threshold in relation to the attribute identified by the user; and Each of the sample images is displayed for labeling by a user, thereby generating the labeled images for the training database.

22. The image enhancement method performed by a device as claimed in claim 19, characterized in that: The attributes of the input image include one or more of the following: scene type, object type in the scene, contrast information, brightness information, edge direction or strength, noise information, blur level, segmentation information, and motion information.

Images