Search

CN-122023583-A - Display dynamic partition image rendering method and system based on content identification

CN122023583ACN 122023583 ACN122023583 ACN 122023583ACN-122023583-A

Abstract

The invention discloses a display dynamic partition image rendering method and a system based on content identification, wherein the method firstly utilizes an RGB data detection module built in a display, combines global identification and 8 multiplied by 6 grid partition identification mechanisms, adopts a lightweight CNN model and a long-short-period memory network to classify content types, adapts to a rendering engine to dynamically adjust a rendering strategy, ensures smooth transition of parameters during content switching through a time delay gradual change algorithm by a transition processing mechanism, and realizes cooperative work of super-resolution sharpening of a text region and wide color gamut mapping of an image region by matching with an edge detection technology. The system also integrates a real-time feedback learning module that allows the user to automatically build a multidimensional preference database containing RGB data, content types, and rendering parameters after marking the unsatisfactory areas. The invention effectively solves the problems of color distortion, insufficient layering, complex operation and the like caused by the fixed rendering mode of the traditional display, and remarkably improves the picture display effect and the user watching experience.

Inventors

  • LI LEI

Assignees

  • 深圳市松冠科技有限公司

Dates

Publication Date
20260512
Application Date
20260128

Claims (10)

  1. 1. The display dynamic partition image rendering method based on content identification is characterized by comprising the following steps of: s1, dynamic content identification: The method comprises the steps of analyzing image characteristics of a display picture in real time through an image RGB data detection and analysis module arranged in a display, simultaneously adopting a global recognition mode, analyzing the image characteristics of a central area of the display picture in real time, adopting a partition recognition mode, analyzing the image characteristics of 8 multiplied by 6 grid partitions of the display picture in real time, and classifying content types through a lightweight CNN model and a sequential analysis model of continuous 5 frames of images; s2, an adaptive rendering engine: According to the result of content type identification in the step S1, a rendering strategy is dynamically adjusted, a dynamic color space mapping algorithm is adopted, when an input signal is sRGB, DCI-P3 color gamut is automatically converted, color restoration errors are kept, when HDR10+ metadata is detected, 12-bit color depth rendering is started, color transition smoothness of a bright field and a dark field is improved, and for a mixed content scene, partition color space independent mapping is adopted; S3, a transitional processing mechanism: When the content types in the step S2 are switched, a time delay gradual change algorithm is adopted, and under a mixed content scene, the text region is independently sharpened according to region segmentation, the video and image region is independently rendered, and at the edge of the region, a blurring algorithm for gradually weakening the color and switching to closing the color enhancement is adopted; s4, real-time quality feedback and self-learning: the user marks the unsatisfactory region through the OSD menu, then the system automatically collects RGB data, content types and rendering parameters of the region, a user preference database is established, and a reinforcement learning algorithm is adopted to dynamically adjust the rendering parameter weight according to the historical preference data.
  2. 2. The method for dynamically partitioning image rendering of display based on content recognition as set forth in claim 1, wherein the text recognition of the content type in step S1 specifically includes the steps of: Firstly, extracting Sobel gradient characteristics of character edges, and representing that edge transition is obvious when the gradient mean value is larger than a threshold value T1; And calculating the dispersion of the picture colors by a color histogram entropy method, namely dividing the picture colors into 16 intervals from small to large as a color histogram according to the RGB gray value, counting the number of pixels occupied by the image in the 16 intervals, calculating the probability value of the number of pixels in the 16 intervals accounting for the total number of pixels, and calculating the entropy value, wherein when the entropy value is smaller than a threshold value T2, the picture colors are concentrated in distribution, the dispersion degree is low, and when the conditions of obvious edge transition and concentrated color distribution are met, the picture is judged to be text.
  3. 3. The method for dynamically partitioning image rendering on a display based on content recognition as set forth in claim 1, wherein in said step S1, when classifying content types by a lightweight CNN model, said lightweight CNN model adopts a depth separable convolution structure, and extracts high-frequency edge features and low-color variance features for text types and continuous tone changes and high-color depth features for image and video types.
  4. 4. The method for dynamically partitioning image rendering of display based on content recognition as set forth in claim 1, wherein the sequential analysis model of continuous 5 frames of image in step S1 adopts a long-short-period memory network structure, probability prediction is performed on the content type of the next frame by analyzing the content type variation trend of the history frame, and loading of the corresponding rendering strategy is triggered in advance when the confidence of the corresponding type prediction reaches a preset high threshold.
  5. 5. The method for rendering dynamic partition images of a display based on content recognition as set forth in claim 1, wherein the dynamic color space mapping algorithm in step S2 is characterized in that when converting sRGB signals into DCI-P3 color gamut, the color mapping lookup table is built and the color gamut compression algorithm is combined to keep the converted colors consistent within the human eye perception range, and when HDR10+ metadata is detected, 12-bit color depth rendering is enabled, and by increasing the number of color quantization stages, the color transition between the bright and dark scenes of sunlight halo and night scene light in a picture is smoother and more natural, and the color banding phenomenon occurring under 8-bit color depth is eliminated.
  6. 6. The method for dynamically partitioning image rendering on display based on content recognition as set forth in claim 1, wherein the time delay gradual change algorithm in step S3 is characterized in that when the content type is switched, a linear interpolation mode is adopted to carry out smooth transition on rendering parameters of brightness, contrast and color, and the transition time length is automatically adjusted according to the content switching type, and the contour of a text region is positioned by region segmentation rendering under a mixed content scene through an edge detection algorithm, a super-resolution sharpening algorithm is independently enabled on the text region, and a wide-color gamut mapping and dynamic contrast optimization are independently enabled on video and image regions.
  7. 7. The method of claim 1, wherein in step S4, when the user marks the unsatisfactory region through the OSD menu, the system synchronously records the coordinate position of the marked region, the frame information of the picture content and the current rendering parameter combination, establishes a multidimensional database containing subjective preferences of the user, and simultaneously, the reinforcement learning algorithm dynamically adjusts the rendering parameter weights under different content types based on the user preference database, the construction state, the action and the rewarding model.
  8. 8. A content recognition-based display dynamic partition image rendering system, employing the content recognition-based display dynamic partition image rendering method of any one of claims 1 to 7, comprising: the image detection module is connected with the image analysis module through a data bus and is used for capturing 8 multiplied by 6 grid partition data of a display picture through an image sensor arranged in the display; The image analysis module is used for receiving RGB data of the image detection module, a light CNN model and a long-term and short-term memory network time sequence analysis model are arranged in the image analysis module, and the image analysis module is connected with the self-adaptive rendering engine through a control signal line; The self-adaptive rendering engine is used for dynamically adjusting rendering parameters according to the content type identification result of the image analysis module and is connected with the transition processing module through a data link; the transition processing module is used for realizing smooth transition of the picture content by adopting a time delay gradual change algorithm when the self-adaptive rendering engine module performs rendering strategy switching; The real-time feedback learning module is used for receiving the coordinates of the unsatisfactory area marked by the user through the OSD menu, collecting RGB data, content types and rendering parameters of the unsatisfactory area and storing the RGB data, the content types and the rendering parameters in the user preference database, and is connected with the self-adaptive rendering engine through a bidirectional data interface.
  9. 9. The dynamic partition image rendering system of a display based on content recognition as set forth in claim 8, wherein the image detection module comprises a global recognition unit and a partition recognition unit, the global recognition unit is used for analyzing image features of a central area of a display screen in real time, the partition recognition unit is used for dividing the screen into 8×6 grid partitions, respectively collecting image data for 48 independent partitions and outputting content type parameters corresponding to the partitions, a lightweight CNN model in the image analysis module adopts a layered feature extraction structure, a bottom network of the layered feature extraction structure extracts high-frequency edge features of text through depth separable convolution, and an edge enhancement algorithm is triggered when character edge transition is detected, and a high-level network of the layered feature extraction structure is used for analyzing continuous tone change features of the image and judging the image and video content through color distribution statistics.
  10. 10. The content recognition based display dynamic partition image rendering system of claim 8, wherein the adaptive rendering engine comprises a color space conversion unit that converts a color gamut into a wide color gamut space through a built-in color mapping lookup table when an input signal is sRGB, and automatically enables a high color depth rendering mode when HDR1+ metadata is detected, and a partition rendering control unit for applying independent color gamut mapping policies to a text region and an image region, respectively, for a mixed content scene.

Description

Display dynamic partition image rendering method and system based on content identification Technical Field The invention relates to the field of display application, in particular to a method and a system for rendering dynamic partition images of a display based on content identification. Background With the increasing abundance and diversity of multimedia content, the user's demand for display picture quality is also increasing. Conventional display rendering techniques often employ a fixed rendering mode, and cannot be dynamically adjusted according to actual requirements of the screen content, so that the screen effect is not optimal. In the use process, a user often needs to manually switch the rendering mode according to the picture content, which not only has complicated operation, but also is difficult to ensure that the picture always keeps at the optimal display effect. Common displays typically employ a fixed rendering mode (e.g., text mode/cinema mode) that requires manual switching by a user and cannot be automatically optimized based on the picture content. Particularly in mixed content scenes (such as web page image-text mixed arrangement), a single rendering mode causes unclear text blurring or color distortion, has weak layering sense, cannot meet the working requirements of display application, and therefore, the method and the system for rendering the dynamic partition image of the display based on content identification are provided. Disclosure of Invention The invention provides a display dynamic partition image rendering method based on content identification, which comprises the following steps: s1, dynamic content identification: The method comprises the steps of analyzing image characteristics of a display picture in real time through an image RGB data detection and analysis module arranged in a display, simultaneously adopting a global recognition mode, analyzing the image characteristics of a central area of the display picture in real time, adopting a partition recognition mode, analyzing the image characteristics of 8 multiplied by 6 grid partitions of the display picture in real time, classifying content types through a lightweight CNN model and a sequential analysis model of continuous 5 frames of images, wherein the content types comprise texts, images and videos, and providing a basis for accurate rendering and improving the picture display effect through analyzing the picture characteristics in real time and classifying the content types; s2, an adaptive rendering engine: According to the result of content type identification in step S1, dynamically adjusting rendering strategy, namely maximizing sharpness by text mode, maximizing color gamut, saturation and dynamic range by image mode, simultaneously adopting dynamic color space mapping algorithm, when input signal is sRGB, automatically converting to DCI-P3 color gamut and keeping color restoration error, when HDR10+ metadata is detected, enabling 12-bit color depth rendering, improving color transition smoothness of bright field and dark field, and aiming at mixed content scene, adopting partition color space independent mapping, namely text area maintaining sRGB color gamut to ensure readability, and video area mapping to wide color gamut and exciting dynamic tone mapping algorithm, optimizing picture quality by dynamically adjusting rendering strategy, making picture clearer and color richer, improving viewing experience; S3, a transitional processing mechanism: When the content types in the step S2 are switched, a time delay gradual change algorithm is adopted, in addition, under a mixed content scene, the text region is independently sharpened according to region division, the video and image region is independently rendered, and at the edge of a region, a blurring algorithm for gradually weakening the color and switching off the color enhancement is adopted, so that smooth transition of a picture is realized, jump is avoided, and the viewing comfort level under the mixed content scene is improved; s4, real-time quality feedback and self-learning: The user marks the unsatisfactory region through the OSD menu, then the system automatically collects RGB data, content types and rendering parameters of the region, a user preference database is established, a reinforcement learning algorithm is adopted, the weight of the rendering parameters is dynamically adjusted according to the historical preference data, the rendering effect is optimized according to user feedback, personalized requirements are met, and user satisfaction is improved. The invention provides a display dynamic partition image rendering system based on content identification, which adopts the display dynamic partition image rendering method based on content identification, and comprises the following steps: The system comprises an image detection module, an image analysis module, an image output module, a self-adaptive rendering engine, a transition proc