Method and device for enhancing image contrast in curved display screen

Abstract

The invention discloses a method and a device for enhancing image contrast in a curved display screen, and belongs to the technical field of display. S2, selecting target brightness; s3, image recognition; s4, calculating an image mean value after each time of histogram adjustment, comparing the image mean value with a preset target value, selecting a parameter with a result closest to a preset target mean value as an optimal adjustment parameter, and selecting the optimal adjustment parameter to process the histogram; s5, performing equalization on the input histogram by using the optimal adjustment parameter; s6, correcting and optimizing the image; according to the method, image recognition is set, image features are extracted, then the features are classified through the features, the classified images are converted into histograms, finally, the category of the images is recognized according to the classified histograms, and finally contrast setting is automatically adjusted according to the image types. Therefore, the best watching experience is provided.

Description

Technical Field

[0001] The present invention belongs to the field of display technology, and in particular relates to a method and a device for enhancing image contrast in a curved display screen. Background Art

[0002] Curved display refers to a display device with a curved panel. First of all, visually, curved display has a better experience than ordinary display, avoiding too large viewing distance at both ends. The curvature of the curved screen can ensure equal distance between the eyes, so that the curved screen can bring a better sensory experience and enhance the user's sense of breadth in visual experience. Curved display gives people a wider field of view, because the edge slightly curved towards the user can be closer to the user, and achieve basically the same viewing angle as the center of the screen. At the same time, the curved screen can make people experience better viewing effects. Contrast enhancement is to stretch or compress the range of brightness values ​​in the image into the brightness display range specified by the display system, thereby improving the contrast of the entire or partial image. Each brightness value in the input image corresponds to a display value of the output image through a certain conversion function. Image contrast enhancement is a common technique in image processing, which aims to improve the visual effect of the image and make the details in the image more clearly visible. By increasing the difference between different gray levels in the image, contrast enhancement can make the image look more vivid and vivid.

[0003] The existing technology has the following problems: 1. Existing curved displays cannot automatically adjust contrast settings according to the type of displayed image, and cannot provide the best viewing experience; 2. When the existing curved display screen equalizes the image histogram, the image is divided into several small areas. When the several small areas are equalized, there is excessive contrast enhancement or excessive noise enhancement, and the quality of the equalized image cannot be further improved. Summary of the invention

[0004] In order to solve the problems raised in the above background technology, the present invention provides a method and device for enhancing the contrast of an image in a curved display screen, which has the characteristics of automatically adjusting the contrast according to the type of displayed image to provide the best viewing experience.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for enhancing the contrast of an image in a curved display screen, comprising the following steps: S1: input image; S2: Select target brightness; S3: Image recognition; S4: Calculate the image mean after each histogram adjustment, compare it with the preset target value, select the parameter whose result is closest to the preset target mean as the best adjustment parameter, and select the best adjustment parameter to process the histogram; S5: Equalize the input histogram using the optimal adjustment parameters; S6: Image correction optimization; S7: Output image.

[0006] Furthermore, image recognition includes the following steps: Feature extraction: extract local features from the input image; Clustering: clustering feature vectors into visual words to form a dictionary; Image representation: Representing an image as a histogram of visual words; Classifier: Classify the scene and then select the best adjustment parameters to process the histogram.

[0007] Furthermore, the image correction optimization includes the following steps: Noise Removal: Replace each pixel value with the median value of its neighboring pixels to reduce noise while preserving image edges and details: Edge enhancement: In image equalization, the image is divided into several small areas, and the edges of the images in several small areas are enhanced by the Soble operator; Color correction: Correct the white balance of an image by adjusting its red, green, and blue channels; Image fusion: Fusion of image processing results at different scales to preserve details at different scales.

[0008] Furthermore, the classifier may classify the images into categories such as movies, games or photos.

[0009] Furthermore, the target brightness is selected by manual input or automatic selection of a default value.

[0010] Furthermore, the feature extraction method uses SIFT or HOG to extract image features.

[0011] Furthermore, after the image is represented as a histogram of visual words, the size of the histogram adjustment function is controlled by a weight parameter, and the histogram is adjusted to different degrees to obtain adjusted histograms under different parameter combinations.

[0012] Furthermore, the scene classification uses a SVM classifier to perform scene classification on the image.

[0013] Furthermore, the clustering uses a K-means algorithm to cluster the feature vectors into visual words to form a dictionary, and the noise removal uses a median filtering method to remove the noise.

[0014] An image contrast enhancement device is also provided, which is used to execute any one of the above-mentioned methods for enhancing the image contrast in a curved display screen.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention sets image recognition, extracts image features, then classifies features by features, converts the classified image into a histogram, and finally identifies the category of the image according to the classified histogram, and finally automatically adjusts the contrast setting according to the image type, thereby providing the best viewing experience.

[0016] 2. The present invention sets image correction optimization and divides the image into several small areas when equalizing the image. When equalizing the several small areas, there is excessive contrast enhancement or excessive noise enhancement. Through noise removal, edge enhancement, color correction and image fusion, the quality of the equalized image is further improved to make it more suitable for specific application requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of the present invention; Figure 2 This is a flowchart of the image recognition process of the present invention; Figure 3 This is a flowchart of the image correction optimization process of the present invention; DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] See also Figure 1-3 The present invention provides the following technical solution: a method and device for enhancing image contrast in a curved display screen, comprising the following steps: S1: input image; S2: Select target brightness; S3: Image recognition; S4: Calculate the image mean after each histogram adjustment, compare it with the preset target value, select the parameter whose result is closest to the preset target mean as the best adjustment parameter, and select the best adjustment parameter to process the histogram; S5: Equalize the input histogram using the optimal adjustment parameters; S6: Image correction optimization; S7: Output image.

[0020] Further in the present invention, image recognition comprises the following steps: Feature extraction: extract local features from the input image; Clustering: clustering feature vectors into visual words to form a dictionary; Image representation: Representing an image as a histogram of visual words; Classifier: Classify the scene and then select the best adjustment parameters to process the histogram.

[0021] By adopting the above technical solution, the contrast setting is automatically adjusted according to the image type to provide the best viewing experience.

[0022] Further in the present invention, the image correction optimization comprises the following steps: Noise Removal: Replace each pixel value with the median value of its neighboring pixels to reduce noise while preserving image edges and details: Edge enhancement: In image equalization, the image is divided into several small areas, and the edges of the images in several small areas are enhanced by the Soble operator; Color correction: Correct the white balance of an image by adjusting its red, green, and blue channels; Image fusion: Fusion of image processing results at different scales to preserve details at different scales.

[0023] By adopting the above technical solutions, the quality of the equalized image is further improved through noise removal, edge enhancement, color correction and image fusion, making it more suitable for specific application requirements.

[0024] Furthermore, in the present invention, the classifier can classify images into categories such as movies, games or photos. The automated process can quickly process a large number of images and improve work efficiency. Based on user interests and behavior patterns, the classifier can help provide more personalized recommendations, such as recommending similar movie posters or game screenshots. Through classification, users can find specific types of content they are interested in more quickly, improving the search experience. By analyzing the popularity of images in different categories, valuable information such as market trends and user preferences can be obtained. By classifying and analyzing image content, companies can make more informed business decisions, such as advertising strategies and product design.

[0025] By adopting the above technical solution, the classifier can classify images into different categories according to image categories.

[0026] Further in the present invention, the target brightness is selected by manual input or automatic selection of a default value.

[0027] By adopting the above technical solution, users can choose the appropriate brightness according to their needs. Users can freely select the target brightness according to their needs and environmental conditions to meet personalized needs. It is suitable for different types of display devices and viewing environments. Users can make adjustments according to actual conditions. The system can automatically adjust the brightness according to current ambient light, display content and other factors to provide a more intelligent user experience. The automatically selected default value can ensure consistent display effects between different devices and users, avoiding problems caused by inconsistent manual settings. It is particularly suitable for scenarios where multiple users share the same device or system. The automatically selected default value can reduce the risk of user misoperation and ensure that the system is always in the best working state. It is particularly suitable for professional application scenarios that have strict requirements on brightness, avoiding problems caused by user misoperation.

[0028] Furthermore, in the present invention, the feature extraction method uses SIFT or HOG to extract image features.

[0029] By adopting the above technical solution, SIFT constructs a Gaussian difference pyramid to find possible feature points. These points are stable local maximum or minimum points at different scales. By fitting a three-dimensional quadratic function to the pixel values ​​around each candidate point, the position and scale of the key point can be determined more accurately, and low-contrast key points and edge responses can be removed. The direction of the key point is calculated based on the direction of the local image gradient, which makes the feature rotation invariant. In a neighborhood around each key point, the gradient size and direction are calculated, and then this information is quantized into a vector to form a descriptor of the key point. Scale and rotation invariance: The SIFT algorithm can detect stable feature points at different scales and rotations. SIFT has the following advantages: illumination invariance: insensitive to changes in image brightness, anti-occlusion and background interference: even if part of the target is occluded or the background is complex, it can maintain good matching performance, good uniqueness: the generated feature vector has high uniqueness and is suitable for fast retrieval in large-scale databases; HOG For the input image, first calculate the gradient strength and direction of each pixel point, divide the image into small connected areas, called cell units, and in each cell unit, construct a histogram according to the gradient direction. Each bin of the histogram represents a specific directional interval. In order to reduce the impact of lighting changes and shadows, the blocks composed of a group of adjacent cell units are usually normalized for contrast. Finally, the histograms of all cell units are connected to form the final feature vector. HOG has the following advantages: Good shape description ability: HOG features have good descriptive capabilities for the shape of objects and are particularly suitable for tasks such as pedestrian detection. They have high computational efficiency: Compared with SIFT, HOG features are simpler and faster to calculate. Robustness to changes in lighting and shadows: Contrast normalization improves robustness to changes in lighting conditions. Local features: HOG focuses on local structural information, which helps improve tolerance to occlusion.

[0030] Furthermore, in the present invention, after the image is represented as a histogram of visual words, the size of the histogram adjustment function is controlled by a weight parameter, and the histogram is adjusted to different degrees to obtain adjusted histograms under different parameter combinations.

[0031] By adopting the above technical solution, the weight of important visual words can be enhanced and the weight of unimportant or common visual words can be reduced through weight parameter adjustment, thereby improving the distinguishability of features. The weight parameters can be dynamically adjusted according to the statistical characteristics of the training data to make the feature representation more robust and adapt to different image data sets and task requirements. Different tasks may require different feature representations. By adjusting the weight parameters, the feature representation can be flexibly optimized to adapt to different application scenarios, such as image retrieval, classification, clustering, etc. The adjusted histogram can better reflect the semantic information of the image, thereby improving the accuracy of image retrieval and classification. By reducing the weight of unimportant visual words, the dimension of the feature vector can be reduced, thereby reducing the computational complexity of subsequent processing.

[0032] Furthermore, in the present invention, scene classification uses a SVM classifier to perform scene classification on images.

[0033] By adopting the above technical solutions, we collect and annotate the training data set, ensure that each image has the correct label to indicate which scene category it belongs to, preprocess the image, such as resizing, graying or color space conversion, etc., in order to unify the input format, and use appropriate feature extraction methods (such as SIFT, HOG or other deep learning methods mentioned above) to extract feature vectors from each image. The quality of feature extraction directly affects the classification effect, so it is very important to select features suitable for the task. The extracted features can be reduced in dimension through PCA (principal component analysis), LDA (linear discriminant analysis) and other technologies to reduce the computational cost and avoid overfitting problems. The extracted feature vectors and the corresponding labels are input into the SVM as a training set, and the appropriate kernel function (such as linear kernel, polynomial kernel, RBF kernel, etc.) is selected, and the parameters (such as C value and γ value) are adjusted to optimize the model performance, and methods such as cross-validation are used. Evaluate the generalization ability of the model to ensure that the model does not overfit the training data. Use an unseen test set to test the trained SVM model and evaluate its accuracy, recall, F1 score and other indicators. Further tune the model parameters or improve the feature extraction method based on the evaluation results. In practical applications, the new image is input into the trained SVM model after the same preprocessing and feature extraction process to obtain the prediction result. SVM maximizes the interval between different categories by finding the optimal hyperplane, thereby achieving efficient and accurate classification. By adjusting the regularization parameters, SVM can effectively control the complexity of the model, avoid overfitting, and ensure good generalization ability. It is suitable for small sample learning: Even when the number of training samples is small, SVM can still perform well. Using the kernel technique, SVM can solve nonlinear classification problems, that is, when the data is not linearly separable, the linear boundary can be found by mapping it to a high-dimensional space.

[0034] Furthermore, in the present invention, the K-means algorithm is used for clustering to cluster the feature vectors into visual words to form a dictionary, and the median filtering method is used for noise removal to remove the noise.

[0035] By adopting the above technical solution, the key point features of each image are extracted, and each key point will generate a feature vector. These feature vectors constitute a large data set. K feature vectors are randomly selected as the initial cluster centers, or a smarter method (such as K-means++) is used to initialize the cluster centers to speed up convergence and improve clustering quality. For each feature vector, its distance to all cluster centers is calculated (Euclidean distance is commonly used), and the feature vector is assigned to the cluster center with the nearest distance. For each cluster, the average value of all feature vectors in it is recalculated and used as the new cluster center. Steps 3 and 4 are repeated until the cluster center no longer changes or the predetermined maximum number of iterations is reached. The final cluster center is the "visual vocabulary". Each cluster center represents a class of similar local features. These visual words together constitute a visual vocabulary table or "dictionary". For new images, local features are also extracted, and then these features are mapped to the nearest visual vocabulary according to the constructed visual vocabulary table. The frequency of each visual vocabulary is counted to form a histogram. This histogram can be used as a global descriptor of the image for subsequent image classification, retrieval and other tasks. The K-means algorithm has the advantages of being simple to implement, high computational efficiency and flexibility. The median filter selects an odd-sized window, such as 3×3, 5×5, etc. The size of the window depends on the noise level and detail requirements of the image. A larger window can better remove noise, but it may blur the details of the image. For each pixel in the image, place a filter window of the selected size with the pixel as the center, obtain the grayscale value or color value of all pixels in the filter window, and sort the pixel values ​​in the window. You can use algorithms such as quick sort and bubble sort to find the middle value after sorting. For a 3×3 window, the middle value is the 5th value; for a 5×5 window, the middle value is the 13th value. Replace the original pixel value at the center of the window with the calculated median value, and repeat the above process for each pixel in the image until the entire image is processed. Median filtering is particularly suitable for removing Salt and pepper noise in the image, because salt and pepper noise is usually isolated points randomly distributed on the image, median filtering can suppress these isolated points well. Compared with linear filters such as mean filtering, median filtering can better retain the edge information of the image while removing noise, avoiding edge blurring caused by excessive smoothing. The calculation of median filtering is relatively simple, and it is easy to implement without complex mathematical operations. Median filtering is highly robust to extreme values ​​(ie noise points) and is not easily affected by a few outliers. Median filtering is not only applicable to grayscale images, but also to color images. Color images can be processed by performing median filtering on each channel separately.

[0036] The working principle and use process of the present invention are as follows: When the method and device for enhancing the contrast of an image in a curved display screen are used, the device is installed in a suitable position, and image recognition is set to extract image features, and then the features are classified by features, and the classified image is converted into a histogram. Finally, the category of the image is identified according to the classified histogram, and finally the contrast setting is automatically adjusted according to the image type, so as to provide the best viewing experience. By setting image correction optimization, the image is divided into several small areas when equalizing the image. When the several small areas are equalized, there is excessive contrast enhancement or excessive noise enhancement. The quality of the equalized image is further improved through noise removal, edge enhancement, color correction and image fusion, so that it is more suitable for specific application requirements.

[0037] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for enhancing image contrast in a curved display screen, characterized in that: The steps include: S1: input image; S2: Select target brightness; S3: Image recognition; S4: Calculate the image mean after each histogram adjustment, compare it with the preset target value, select the parameter whose result is closest to the preset target mean as the best adjustment parameter, and select the best adjustment parameter to process the histogram; S5: Equalize the input histogram using the optimal adjustment parameters; S6: Image correction optimization; S7: Output image.

2. The method for enhancing image contrast in a curved display screen according to claim 1, characterized in that: Image recognition includes the following steps: Feature extraction: extract local features from the input image; Clustering: clustering feature vectors into visual words to form a dictionary; Image representation: Representing an image as a histogram of visual words; Classifier: Classify the scene and then select the best adjustment parameters to process the histogram.

3. The method for enhancing image contrast in a curved display screen according to claim 1, characterized in that: Image correction optimization includes the following steps: Noise Removal: Replace each pixel value with the median value of its neighboring pixels to reduce noise while preserving image edges and details: Edge enhancement: In image equalization, the image is divided into several small areas, and the edges of the images in several small areas are enhanced by the Soble operator; Color correction: Correct the white balance of an image by adjusting its red, green, and blue channels; Image fusion: Fusion of image processing results at different scales to preserve details at different scales.

4. The method for enhancing image contrast in a curved display screen according to claim 2, characterized in that: The classifier may classify images into movie, game, or photo categories.

5. The method for enhancing image contrast in a curved display screen according to claim 2, characterized in that: The target brightness is selected by manual input or automatic selection of a default value.

6. The method for enhancing image contrast in a curved display screen according to claim 2, characterized in that: The feature extraction method uses SIFT or HOG to extract image features.

7. The method for enhancing image contrast in a curved display screen according to claim 2, characterized in that: After the image is represented as a histogram of visual words, the size of the histogram adjustment function is controlled by a weight parameter, and the histogram is adjusted to different degrees to obtain adjusted histograms under different parameter combinations.

8. The method for enhancing image contrast in a curved display screen according to claim 2, characterized in that: The scene classification uses a SVM classifier to perform scene classification on an image.

9. The method for enhancing image contrast in a curved display screen according to claim 2, characterized in that: The clustering uses a K-means algorithm to cluster feature vectors into visual words to form a dictionary, and the noise removal uses a median filtering method to remove noise.

10. An image contrast enhancement device, characterized in that: Used to execute the method for enhancing image contrast in a curved display screen as described in any one of claims 1-9.

Images