CN-122023871-A - Digital photo frame image intelligent processing method and system based on deep learning

CN122023871ACN 122023871 ACN122023871 ACN 122023871ACN-122023871-A

Abstract

The invention relates to the field of image processing and discloses a digital photo frame image intelligent processing method and a digital photo frame image intelligent processing system based on deep learning, wherein the method comprises the steps of obtaining original image data and preprocessing to obtain a preprocessed image; the method comprises the steps of utilizing a convolutional neural network to extract image feature vectors, analyzing complexity and classifying based on the image feature vectors to obtain an image classification result, determining an adaptive parameter set according to the classification result, obtaining light ray data, classifying the image feature vectors based on the light ray data, extracting semantic feature vectors and adjusting weights of the semantic feature vectors to generate dynamic strategy vectors, carrying out refinement treatment on the dynamic strategy vectors based on the light ray data to obtain a refinement strategy set, adjusting images according to the refinement strategy set, and carrying out association comparison iterative optimization to obtain a final optimized image. The invention realizes self-adaptive and high-quality image display optimization by constructing a multidimensional feature space and introducing a closed loop iteration optimization mechanism.

Inventors

NIE DA
ZHANG JIAN

Assignees

深圳市一诺成电子有限公司

Dates

Publication Date: 20260512
Application Date: 20251223

Claims (10)

1. A digital photo frame image intelligent processing method based on deep learning is characterized by comprising the following steps: Acquiring original image data and preprocessing to obtain a preprocessed image; Extracting an image feature vector from the preprocessed image by using a convolutional neural network; analyzing the complexity of the image based on the image feature vector and classifying the image to obtain an image classification result; Acquiring display demand data of a current digital photo frame according to the image classification result, and determining an adaptive parameter set matched with the image feature vector; Acquiring light data, classifying the image feature vectors based on the light data, extracting semantic feature vectors based on classification results, calculating the matching degree of the classification results and the semantic feature vectors, adjusting the weight of the semantic feature vectors, and generating optimized dynamic strategy vectors; carrying out refinement treatment on the dynamic policy vector based on the light ray data to obtain a refinement policy set; adjusting the preprocessed image according to the refinement strategy set to obtain an adjusted image; and carrying out association comparison on the adjusted image and the image feature vector, and optimizing the adjusted image according to an association comparison result to obtain a final optimized image.
2. The intelligent processing method for digital photo frame images based on deep learning according to claim 1, wherein the steps of obtaining original image data and preprocessing the original image data to obtain a preprocessed image include: Acquiring original image data; performing gray level conversion and noise reduction on the original image data to obtain a first image; And based on the first image, performing edge detection and image segmentation operations to obtain a preprocessed image divided into a plurality of areas.
3. The intelligent processing method for digital photo frame images based on deep learning according to claim 1, wherein the extracting the image feature vector from the preprocessed image by using a convolutional neural network comprises: When the number of the areas determined by the preprocessed image exceeds a preset area number threshold, extracting features of each area through a convolutional neural network to obtain feature vectors of color distribution, texture details and semantic content, and obtaining a multidimensional feature set; and performing dimension reduction processing on the multi-dimensional feature set by adopting a principal component analysis method to obtain an image feature vector.
4. The intelligent processing method for digital photo frame images based on deep learning according to claim 1, wherein the analyzing image complexity based on the image feature vector and classifying the images to obtain image classification results comprises: Extracting statistical characteristics of texture detail components in the image feature vector by using a gray level co-occurrence matrix, and calculating an entropy value to obtain first complexity; When the first complexity exceeds a preset complexity threshold, grouping the image feature vectors by adopting K-means clustering to obtain a grouping feature set; calculating the ratio of the internal variance to the external variance of each feature cluster in the grouping feature set to obtain a variance ratio set; and when at least one ratio in the variance ratio set exceeds a preset variance ratio threshold, acquiring boundary complexity through boundary point analysis, and determining an image classification result.
5. The intelligent processing method for digital photo frames based on deep learning according to claim 1, wherein the step of obtaining display requirement data of a current digital photo frame according to the image classification result, and determining an adaptive parameter set matched with the image feature vector comprises the steps of: Extracting brightness distribution characteristics by adopting a gray level histogram method according to the image classification result, and calculating brightness mean value and variance to obtain brightness distribution description; when the variance in the brightness distribution description exceeds a preset brightness variance threshold, converting an image from an RGB color space to a CIELAB color space to obtain a color gamut range parameter; Grouping the image feature vectors through K-means clustering, calculating the pixel density mean value of each grouping, and determining a display parameter set matched with the resolution; And adjusting the contrast and color saturation of the image by adopting self-adaptive mapping according to the color gamut range parameters and the display parameter set to obtain an adaptive parameter set.
6. The intelligent processing method for digital photo frame image based on deep learning according to claim 1, wherein the obtaining light data, classifying the image feature vector based on the light data, extracting a semantic feature vector based on a classification result, calculating a matching degree between the classification result and the semantic feature vector, adjusting weights of the semantic feature vector, and generating an optimized dynamic policy vector comprises: acquiring light intensity data from an ambient light sensor, and smoothing the light intensity data by adopting a mean value filtering method to obtain a light characteristic vector; When the average value of the light characteristic vectors exceeds a preset light intensity threshold value, classifying the image characteristic vectors through a support vector machine to obtain a first classification result; Extracting semantic characteristic vectors corresponding to the first classification result from a pre-established semantic characteristic library, and calculating the matching degree of the semantic characteristic vectors and the first classification result; When the matching degree is lower than a preset matching degree threshold value, the weight of the semantic characteristic vector is adjusted by adopting a gradient descent method, and an optimized dynamic strategy vector is generated.
7. The intelligent processing method for digital photo frame images based on deep learning according to claim 1, wherein the refining the dynamic policy vector based on the ray data to obtain a refined policy set comprises: selecting a subset related to scene adaptation from the dynamic policy vectors; Clustering and grouping the subsets to obtain a grouping set; Acquiring texture capturing information, and fusing the texture capturing information and the ray data according to the grouping set to generate a fused feature vector; And classifying the fusion feature vectors by adopting a decision tree to generate a refinement strategy set.
8. The intelligent processing method for digital photo frame images based on deep learning according to claim 1, wherein the adjusting the preprocessed image according to the refinement policy set to obtain an adjusted image comprises: acquiring color saturation and brightness characteristics from the preprocessed image, and carrying out normalization processing to obtain a normalized image; extracting edge features and texture features according to the normalized image, and fusing to obtain a fused feature set; When the average value of the color saturation of the fusion feature set is lower than a preset saturation threshold value, adopting self-adaptive histogram enhancement to adjust the color saturation; Determining brightness adjustment parameters according to the refinement strategy set; And combining the brightness adjustment parameter and the adjusted color saturation to generate an adjusted image.
9. The intelligent processing method of digital photo frame image based on deep learning according to claim 1, wherein the performing association comparison on the adjusted image and the image feature vector, and optimizing the adjusted image according to the association comparison result to obtain a final optimized image, comprises: Calculating a comparison value of the adjusted image and the image feature vector; When the comparison value is lower than a preset comparison threshold value, a first comparison result is generated; when the comparison value is not lower than the preset comparison threshold value, a second comparison result is generated; When the comparison result is the first comparison result, iteratively updating the dynamic strategy vector, and returning to the step of refining the dynamic strategy vector based on the light ray data; and when the comparison result is the second comparison result, outputting the adjusted image as a final optimized image.
10. An intelligent digital photo frame image processing system based on deep learning is characterized by comprising: the preprocessing module is used for acquiring original image data and preprocessing the original image data to obtain a preprocessed image; The feature extraction module is used for extracting image feature vectors from the preprocessed image by using a convolutional neural network; The classification module is used for analyzing the image complexity based on the image feature vector and classifying the image to obtain an image classification result; The parameter determining module is used for acquiring display demand data of the current digital photo frame according to the image classification result and determining an adaptive parameter set matched with the image feature vector; The dynamic policy generation module is used for acquiring light data, classifying the image feature vectors based on the light data, extracting semantic feature vectors based on classification results, calculating the matching degree of the classification results and the semantic feature vectors, adjusting the weight of the semantic feature vectors and generating optimized dynamic policy vectors; the policy refinement module is used for refining the dynamic policy vector based on the light data to obtain a refinement policy set; the image adjustment module is used for adjusting the preprocessed image according to the refinement strategy set to obtain an adjusted image; and the final output module is used for carrying out association comparison on the adjusted image and the image feature vector, and optimizing the adjusted image according to an association comparison result to obtain a final optimized image.

Description

Digital photo frame image intelligent processing method and system based on deep learning Technical Field The invention relates to the field of image processing, in particular to an intelligent processing method and system for digital photo frame images based on deep learning. Background Currently, as an important tool for modern home and personalized display, the image processing technology of the digital photo frame directly influences the visual experience and the use feeling of a user. With the rapid development of intelligent image processing technology, the digital photo frame can realize automatic optimization through analysis of image content so as to meet diversified aesthetic requirements and requirements of different application scenes, and the core technology relates to a plurality of technical fields such as image enhancement, image recognition and the like. In one prior art, the image processing method of the digital photo frame mainly relies on feature analysis with a single dimension, for example, only adjusting the color or brightness of the image. The method improves the visual effect of the image to a certain extent, but often cannot fully capture the diversified characteristics of the image, including texture details, content semantics and style characteristics, when processing complex images. It can be deduced from this that the image processing method of single dimension or fixed strategy is difficult to meet the requirements of multiple scenes and multiple devices at the same time. In a typical implementation of the prior art, the feature analysis dimension is relatively single, and it is difficult to effectively integrate multidimensional features and dynamically map the multidimensional features to an optimal processing strategy, so that the problem that the images are easy to distort or have inconsistent styles after being processed is caused, and finally, the problem of poor display effect is caused. Disclosure of Invention The invention provides a digital photo frame image intelligent processing method and system based on deep learning, which are used for solving the problem that the prior art cannot realize self-adaptive image optimization in a complex scene due to the dependence on single-dimension feature analysis, so that the display effect is poor. In order to solve the above technical problems, the present invention provides a digital photo frame image intelligent processing method based on deep learning, including: Acquiring original image data and preprocessing to obtain a preprocessed image; Extracting an image feature vector from the preprocessed image by using a convolutional neural network; analyzing the complexity of the image based on the image feature vector and classifying the image to obtain an image classification result; Acquiring display demand data of a current digital photo frame according to the image classification result, and determining an adaptive parameter set matched with the image feature vector; Acquiring light data, classifying the image feature vectors based on the light data, extracting semantic feature vectors based on classification results, calculating the matching degree of the classification results and the semantic feature vectors, adjusting the weight of the semantic feature vectors, and generating optimized dynamic strategy vectors; The dynamic strategy vector is refined based on the light ray data to obtain a refined strategy set, and the preprocessed image is adjusted according to the refined strategy set to obtain an adjusted image; and carrying out association comparison on the adjusted image and the image feature vector, and optimizing the adjusted image according to an association comparison result to obtain a final optimized image. In an alternative embodiment, the acquiring the original image data and preprocessing to obtain a preprocessed image includes: Performing gray conversion and noise reduction on the original image data to obtain a first image; And based on the first image, performing edge detection and image segmentation operations to obtain a preprocessed image divided into a plurality of areas. In an alternative embodiment, the extracting the image feature vector from the preprocessed image by using a convolutional neural network includes: When the number of the areas determined by the preprocessed image exceeds a preset area number threshold, extracting features of each area through a convolutional neural network to obtain feature vectors of color distribution, texture details and semantic content, and obtaining a multidimensional feature set; and performing dimension reduction processing on the multi-dimensional feature set by adopting a principal component analysis method to obtain an image feature vector. In an optional embodiment, the analyzing the image complexity based on the image feature vector and classifying the image to obtain an image classification result includes: Extracting statistical characteristics of