Digital content making method based on image stitching

Abstract

The invention provides a digital content making method based on image stitching, and relates to the technical field of image data processing, image acquisition is carried out through an image acquisition device, the acquired images comprise images shot at a fixed angle, and the overlapping part of the acquired images reaches a preset overlapping threshold value; performing distortion correction on the acquired image to obtain a corrected image; performing color adjustment on the corrected image to obtain a corrected color image; and extracting feature points from the corrected color image by using a scale invariant feature transformation algorithm, and splicing the collected images according to the feature points to obtain a spliced image. According to the method, the technical problem that the image splicing quality and efficiency are affected due to inaccurate feature point matching and large calculation amount caused by complex scenes and illumination changes in an existing scale-invariant feature transformation algorithm is solved, and the technical effects of accurately and efficiently splicing the image and improving the digital content quality are achieved.

Description

Technical Field

[0001] The present application relates to the technical field of image data processing, and in particular to a method for producing digital content based on image splicing. Background Art

[0002] In today's digital age, digital content production has become an important part of many fields such as media, entertainment, and scientific research. It involves the entire conversion process from raw data to final presentation. In the field of digital content production, image stitching technology has become a key means, especially in creating high-resolution, wide-angle or panoramic images. However, the existing image stitching technology based on scale-invariant feature transformation algorithm still has some shortcomings in practical applications. The main problem lies in the accuracy and efficiency of feature point matching. In complex scenes or when the illumination changes greatly, the scale-invariant feature transformation algorithm may produce incorrect feature point matching, resulting in misalignment and distortion of the stitched image, affecting the overall stitching quality. In addition, the computational complexity of the scale-invariant feature transformation algorithm is high. When processing a large number of images or high-resolution images, the feature point matching process takes a long time, which reduces the stitching efficiency. Summary of the invention

[0003] The present application provides a digital content production method based on image stitching, which solves the technical problem that when using a scale-invariant feature transformation algorithm for image stitching, due to complex scenes and lighting changes, feature point matching is inaccurate and the amount of calculation is large, thereby affecting the quality and efficiency of image stitching, thereby achieving the technical effect of accurately and efficiently stitching images and improving the quality of digital content.

[0004] In view of the above problems, the present application provides a digital content production method based on image stitching, the method comprising: performing image acquisition through an image acquisition device, the acquired images obtained include images shot at a fixed angle, and the overlapping parts of the acquired images reach a preset overlapping threshold; performing distortion correction on the acquired images to obtain a corrected image; performing color adjustment on the corrected image to obtain a corrected color image; using a scale-invariant feature transformation algorithm to extract feature points from the corrected color image, and stitching the acquired images according to the feature points to obtain a stitched image.

[0005] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0006] Image acquisition is performed through an image acquisition device, and the acquired images include images shot at a fixed angle. The overlapping parts of the acquired images reach a preset overlapping threshold, ensuring that there is sufficient consistency and overlapping area between images, providing a stable basis for subsequent feature point matching, and reducing stitching errors caused by inconsistent shooting angles. The acquired images are subjected to distortion correction to obtain a corrected image, eliminating or reducing the distortion caused by the lens and shooting conditions, improving the authenticity of the image, and reducing image dislocation caused by distortion during the stitching process. The corrected images are subjected to color adjustment to obtain a corrected color image, eliminating the color difference between different images. The stitched images are made more visually consistent, reducing visual defects caused by color differences. Feature points are extracted from the corrected color image using a scale-invariant feature transformation algorithm, and the acquired images are stitched according to the feature points to obtain a stitched image.

[0007] To summarize, the present application optimizes image acquisition by fixed-angle shooting and pre-setting overlap thresholds, pre-processes images through distortion correction and color adjustment, and finally uses a scale-invariant feature transformation algorithm for feature point extraction and stitching, thereby achieving high-precision and high-efficiency image stitching, and achieving the effect of improving the accuracy of stitched images and the quality of digital content.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 A flowchart of a method for producing digital content based on image stitching provided in an embodiment of the present application;

[0010] Figure 2 A schematic diagram of a process of stitching captured images according to feature points in a digital content production method based on image stitching provided in an embodiment of the present application;

[0011] Figure 3 A schematic diagram of a process for eliminating visual differences at image seams of fused images through feathering stitching in a digital content production method based on image stitching provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a digital content production method based on image stitching, optimize the image acquisition method and preprocessing steps, and solve the technical problem that when using a scale-invariant feature transformation algorithm for image stitching, due to complex scenes and lighting changes, feature point matching is inaccurate and the amount of calculation is large, thereby affecting the quality and efficiency of image stitching, thereby achieving the technical effect of accurately and efficiently stitching images and improving the quality of digital content.

[0013] like Figure 1 As shown, the embodiment of the present application provides a digital content production method based on image splicing, the method comprising:

[0014] The image acquisition device is used to acquire images, and the acquired images include fixed-angle photographed images, and the overlapping parts of the acquired images reach a preset overlapping threshold.

[0015] Specifically, image acquisition devices refer to hardware devices used to capture images, such as digital cameras, cameras mounted on drones, mobile phone cameras, etc. These devices can capture scenes in the real world and convert them into digital images for subsequent processing. Fixed-angle shooting means that during the shooting process, the position or viewing angle of the camera remains unchanged, and images in different directions are acquired only by rotation or translation. For example, when shooting a 360-degree panorama, the camera can be fixed on a tripod and then rotated around the central axis, taking a photo after each rotation of a certain angle. In order to ensure that multiple images can be stitched correctly, a certain proportion of overlapping areas needs to be retained between adjacent images. The preset overlap threshold refers to the minimum overlap ratio set manually during the image acquisition stage, usually 30% to 50%. This ensures that there is enough information for image alignment and stitching during the stitching process.

[0016] In the production of digital content based on image stitching, image acquisition is the first and most important step. First, choose a suitable image acquisition device, for example, use a digital camera with high resolution and wide-angle lens, or a drone equipped with a high-quality camera to adapt to different application scenarios and shooting needs. When collecting images, a fixed-angle shooting method is adopted. For example, the camera can be mounted on a tripod, the shooting range and angle can be set, and then the camera can be moved in a certain order and direction. A new image is taken after each movement, so that each image will have a certain overlap area with the adjacent image. This overlapping area needs to reach a preset overlap threshold. For example, when taking a panoramic view of a building, each photo is overlapped by 40% to ensure that there are enough common features between the images for matching and alignment during subsequent stitching. This step is usually assisted by professional panoramic photography software or the panoramic mode of the camera, such as the Photomerge function of Adobe Photoshop or the panoramic mode of the mobile phone camera. These tools can help users control the shooting angle and overlapping area more accurately, so that image stitching and fusion can be achieved more easily during post-processing. Through this method, it can be ensured that the acquired image sequence is spatially continuous, providing the necessary basis for subsequent image stitching.

[0017] Performing distortion correction on the collected image to obtain a corrected image.

[0018] Specifically, distortion correction refers to the process of correcting the geometric deformation in the image caused by lens characteristics or shooting conditions to restore the true geometric shape of the image. The image obtained after distortion correction of the acquired image is called a corrected image, and its geometric shape is closer to the real scene. Distortion correction of the acquired image is an important step in image preprocessing, which ensures the authenticity and accuracy of the stitched image. This process first requires identifying the type of distortion in the image. Common types of distortion include barrel distortion, that is, the image edge expands outward, and pincushion distortion, that is, the image edge shrinks inward. These distortions are usually caused by the optical characteristics of the lens. The steps to achieve distortion correction include: first, according to the characteristics of the lens and the type of distortion, a mathematical model is established to describe the distortion. Common models include polynomial models, fisheye models, etc. The specific parameters of the distortion model are estimated by analyzing image feature points or using known distortion parameters. Using the estimated distortion parameters, each pixel in the original image is mapped to a new position according to the distortion model to eliminate the distortion. This process is usually completed using image processing software or specialized distortion correction tools, such as Adobe Photoshop, ImageMagick, or the open source computer vision library OpenCV. These tools provide distortion correction algorithms and interfaces, and users only need to enter the distortion coefficient or use the built-in lens profile to automatically perform distortion correction.

[0019] For example, for images taken with a fisheye lens, OpenCV is used for distortion correction, and the image is corrected by inputting the corresponding distortion coefficients to make it present the correct geometric shape. OpenCV is an open source computer vision and machine learning library that provides an API for distortion correction. Users can correct the distortion by reading the image, loading the distortion coefficients, and then calling the undistort function. When shooting a building with a wide-angle lens, the straight lines of the building may appear curved due to the barrel distortion of the lens. The "Lens Correction" filter in Adobe Photoshop can automatically analyze the distortion type of the image and correct it. The user only needs to select the camera and lens model, and the software can automatically load the corresponding distortion configuration for distortion correction.

[0020] Through the above steps, the acquired images can be effectively distorted and corrected, thereby improving the image quality and the accuracy of subsequent processing, making the stitched images more realistic and accurate.

[0021] Performing color adjustment on the corrected image to obtain a corrected color image.

[0022] Specifically, color adjustment refers to the process of calibrating and optimizing the colors in the calibration image to ensure color consistency and visual effects between or within images. Color adjustment includes adjustments to color balance, contrast, brightness, saturation, and other aspects. The calibrated image after color adjustment is called a calibrated color image.

[0023] The colors of the captured images may vary due to factors such as lighting, white balance settings, and camera sensor characteristics. If these differences are not corrected, they will produce obvious color faults when the images are stitched together, affecting the visual experience. Color adjustment of the calibrated images is to solve the color difference problem that may exist between images and ensure that the stitched images are more visually coherent and natural. This process usually includes several steps: color analysis, color correction, and color matching: First, analyze the color distribution of the calibrated image and determine the color parameters that need to be adjusted. According to the results of the color analysis, adjust the image's color balance, brightness, contrast and other attributes. Adjust the color channels such as red, green, blue (RGB) or cyan, magenta, and yellow (CMY) in the image to achieve color balance and consistency. Adjust the brightness and contrast of the image to make the image look clearer and more natural as a whole. Adjust the saturation of the image color to enhance or reduce the vividness of the color to achieve a better visual effect. Correct the color temperature in the image so that white objects in the image appear white under various lighting conditions, thereby ensuring color accuracy. If multiple images are involved in stitching, it is also necessary to ensure color matching between different images to reduce stitching marks. This process can be achieved through a variety of image processing software, such as Adobe Lightroom, Adobe Photoshop, Capture One, etc.

[0024] For example, in Adobe Photoshop, you can use the "color balance" tool to adjust the color cast of an image, such as correcting color inconsistencies caused by different lighting conditions. By adjusting the red, green, and blue channels, you can make the color of the image more natural. Using the "tone curve" function of Adobe Lightroom, you can fine-tune the brightness and contrast of the image to make the overall tone of the image more harmonious, and when stitching panoramic images, you can ensure that different images look like a continuous photo after stitching. In OpenCV, you can use the cvtColor function to convert the color space of an image, such as from BGR to Lab color space, then perform color adjustments in Lab space, and finally convert back to BGR space. This method can automate the color adjustment of a large number of images.

[0025] Through the above steps, the color of the corrected image can be effectively adjusted to reduce visual defects caused by color differences, thereby ensuring the color consistency between images, providing high-quality image materials for subsequent image stitching and other processing, and significantly improving the final effect of image stitching.

[0026] A scale-invariant feature transformation algorithm is used to extract feature points from the corrected color image, and the collected images are spliced ​​according to the feature points to obtain a spliced ​​image.

[0027] Specifically, the scale-invariant feature transform algorithm is a computer vision algorithm used to detect and describe local feature points in an image. These feature points are points with unique properties in the image, such as corner points, edge points, etc., and are invariant to image scaling, rotation, and even lighting changes. Common scale-invariant feature transform algorithms include SIFT, SURF, and ORB.

[0028] A series of feature points are automatically detected in each color-corrected image using a scale-invariant feature transform algorithm. These feature points are scale-invariant and remain stable even when the image is scaled or rotated. For each detected feature point, the scale-invariant feature transform algorithm generates a feature descriptor, which is a vector that contains the gradient direction and intensity information of the area around the feature point. The feature point descriptors between different images are matched to find the most similar point pairs. This process can identify the corresponding parts of the same object or scene in different images. Based on the matched feature point pairs, the geometric transformation model between the images, such as the homography matrix, is calculated, and then the images are transformed to the same coordinate system for seamless stitching. This process is usually implemented using computer vision and image processing software such as OpenCV and Photoshop. Through the above process, a large, high-resolution stitched image can be finally obtained, which combines the information of multiple original images while maintaining the coherence and visual consistency of the scene.

[0029] Furthermore, the embodiment of the present application splices the collected images according to the feature points to obtain a spliced ​​image, and the method also includes:

[0030] Calculate the calculation accuracy of the homography matrix; optimize the homography matrix according to the calculation accuracy to obtain an optimized combination; monitor the registration error in real time through the optimized combination, correct the registration error using a feedback loop, and obtain a registration result; evaluate the registration quality of the registration result, and if the registration threshold is met, retain the homography matrix corresponding to the optimized combination.

[0031] Specifically, the homography matrix is ​​a matrix used to describe the geometric transformation between images. In particular, in image stitching, it can represent the perspective projection relationship from one image to another. The homography matrix is ​​usually a 3x3 matrix. By solving this matrix, the alignment between images can be achieved. The calculation accuracy refers to the accuracy and stability of the algorithm in matching feature points in the process of solving the homography matrix. The calculation accuracy directly affects the quality of image stitching, including the alignment accuracy and stitching effect of the image. The registration error refers to the deviation between the actual matching position and the ideal position between the images in the image registration process. The size of the registration error reflects the accuracy of image registration and is an important indicator for evaluating the quality of image stitching. The feedback loop is used to monitor the registration error in real time and adjust the homography matrix according to the error size to achieve dynamic optimization. The registration threshold is a preset indicator used to determine whether the quality of image registration meets the requirements. If the registration error of the registration result is less than the registration threshold, it means that the image registration quality meets the requirements and the current homography matrix can be retained.

[0032] The homography matrix is ​​solved by matching feature points. The calculated homography matrix needs to be evaluated for accuracy. This can be done by transforming one image into the space of another image through the homography matrix, and then matching the feature points in the transformed image with the feature points in the other image. The error between the matched feature points is calculated, and this error can be used to evaluate the accuracy of the homography matrix. The calculated homography matrix may have certain inaccuracies due to factors such as noise and feature point matching errors. Therefore, the homography matrix needs to be further optimized. This is usually achieved through iterative optimization algorithms such as gradient descent and Levenberg-Marquardt algorithm. In each iteration, the image registration error is calculated by adjusting the parameters of the homography matrix, and the feedback loop is used to dynamically adjust until the registration error is minimized. During the entire optimization process, the registration error is monitored in real time and dynamically adjusted using the feedback loop, which can effectively improve the quality of image registration. Once the registration error is lower than the preset registration threshold, it means that the image registration quality meets the requirements. The homography matrix at this time is the optimized result and can be used for the final stitching of the image.

[0033] Further, such as Figure 2 As shown, the embodiment of the present application splices the collected images according to the feature points to obtain a spliced ​​image, including:

[0034] A homography matrix is ​​calculated based on the feature points, and a transformation relationship of the acquired image is obtained based on the description of the homography matrix; a weighted average of the acquired images is performed according to the transformation relationship to obtain a fused image; and visual differences at image seams of the fused image are eliminated through feathering stitching to obtain the stitched image.

[0035] Specifically, a fused image is an image obtained by weighted averaging the pixel values ​​of multiple images, which combines the information of multiple images. Feathering stitching refers to applying a blur effect at the seams of image stitching to reduce visual differences and make the stitched image more natural.

[0036] Use SIFT, SURF or other feature point detection algorithms to detect feature points in each image and calculate the correspondence between them. Then, using the matched feature point pairs, use the least squares method or RANSAC algorithm to solve the homography matrix and calculate the homography matrix between each image. Based on the calculated homography matrix, describe the geometric transformation relationship between each image. Including parameters such as rotation, scaling, and translation. Apply the homography matrix to each image and align the images through perspective transformation so that they overlap in the stitching area. After the images are aligned, the pixel values ​​of each image in the overlapping area are weighted averaged to generate a fused image. The weighted average can be based on image features such as brightness and clarity, or simply use linear blending to ensure a smooth transition at the splicing. The weight of the weighted average can be determined based on the strength of the feature point matching. The higher the matching strength, the greater the weight. Finally, use the feathering technique to process the splicing of the fused image. By gradually adjusting the transparency of the edge pixels, the image edges are naturally blended, the splicing marks are eliminated, and the final splicing image is obtained.

[0037] Preferably, the embodiment of the present application eliminates visual differences at image seams of the fused image by feathering stitching to obtain the stitched image, and the method further includes:

[0038] The feature point algorithm is spliced ​​according to the feature point scene requirements; the feature point detection parameters are adjusted according to the feature point algorithm; the detection impact on the detection result is calculated according to the feature point detection parameters; the feature points are preliminarily filtered according to the detection impact to remove the wrong splicing points to generate feature points.

[0039] Specifically, feature point detection parameters refer to parameters that need to be adjusted during feature point detection, such as detection threshold, neighborhood size, etc. Detection impact refers to the impact of feature point detection parameters on detection results, including the number, location, and quality of feature points.

[0040] First, according to the scene requirements of the stitched images, such as image size, resolution, lighting conditions, and processing speed requirements, the most appropriate feature point detection algorithm is selected. For example, for scenes that require stable detection under large-scale changes, the SIFT algorithm may be selected; while for real-time applications that require high-speed processing, the ORB algorithm may be selected. According to the selected feature point algorithm, adjust the feature point detection parameters to optimize the detection results of the feature points. This includes adjusting the threshold, corner detection sensitivity, etc., to ensure that the detected feature points are both stable and rich enough to support the subsequent stitching process. For example, adjusting the detection threshold can control the number of feature points, and adjusting the neighborhood size can control the distribution of feature points. By adjusting the parameters and observing the changes in the detection results, including the number, location, and quality of feature points. Evaluate the impact of the parameters on the detection effect by comparing the detection results before and after the adjustment. After the feature points are detected, preliminary filtering is required to remove those mis-stitched points caused by noise, lighting changes, or algorithm misdetection. This is usually achieved by checking the distance, angle, or descriptor similarity between feature points to ensure that only truly matching point pairs are retained. Through these steps, more accurate feature points that better meet the scene requirements can be obtained, thereby improving the quality of the stitched image.

[0041] Preferably, Figure 3 As shown, the embodiment of the present application eliminates the visual difference at the image seam of the fused image by feathering stitching to obtain the stitched image, including:

[0042] The fused image is cropped to remove the distorted part of the image edge to obtain a cropped image; the cropped image is sharpened to obtain a sharpened image; the noise of the sharpened image is reduced by noise removal to obtain a denoised image; the color of the denoised image is enhanced according to image requirements to obtain an enhanced image; the enhanced image is format converted to obtain the stitched image.

[0043] Specifically, since edge distortion may occur during the stitching process, the fused image needs to be cropped. By setting the cropping area, the distorted part of the edge is removed to ensure that the edge of the cropped image is smooth. The image obtained after cropping the fused image is called a cropped image. In order to improve the clarity of the image, the cropped image is sharpened to enhance the edges and details of the image, making the objects and scenes in the image clearer. The image after sharpening is called a sharpened image. Sharpening may introduce noise, so it is necessary to use a noise removal algorithm, such as Gaussian noise removal, median filtering, etc. to remove noise from the sharpened image to obtain a denoised image to improve the image quality. According to the application requirements of the image, the denoised image is color enhanced, and color enhancement techniques such as histogram equalization, color space conversion or adaptive contrast enhancement algorithm are used to adjust the color performance of the denoised image to improve the color effect of the image. The denoised image after color enhancement is called an enhanced image. According to the output requirements, the enhanced image is converted into the required format to obtain the final stitched image. Through these processing steps, high-quality stitched images can be obtained, and their quality and visual effects are effectively improved.

[0044] Preferably, Figure 3 As shown, the embodiment of the present application eliminates the visual difference at the image seam of the fused image by feathering stitching to obtain the stitched image, and also includes:

[0045] Detect and remove dynamic blur in the stitched image to obtain a deblurred image; enhance the pixels of the deblurred image using a high dynamic range to obtain a high dynamic range image; perform local noise reduction on the high dynamic range image according to the image content to obtain a local noise reduction image; use an adaptive filter to correct the local contrast of the local noise reduction image to obtain a corrected image; perform geometric correction on the corrected image to obtain the stitched image.

[0046] Specifically, motion blur refers to blur caused by camera movement or object movement. High dynamic range (HDR) enhancement is an image processing technology that expands the brightness range of an image by merging multiple images with different exposures so that details in both highlight and shadow areas can be preserved. Local noise reduction refers to noise reduction in specific areas of an image to preserve important details. Adaptive filters refer to filters that automatically adjust filtering parameters based on local changes in image content.

[0047] First, use edge detection algorithms, such as Canny edge detection, and texture analysis methods to detect blurred areas in the image. Then, estimate the direction and degree of blur, and remove the blur through deconvolution or sharpening filters, such as inverse filtering or Wiener filtering, to obtain a deblurred image. Use an HDR algorithm to convert the deblurred image into a high dynamic range image to better display the details and dynamic range of the image. This can be achieved through HDR synthesis software such as Photoshop's HDR function. While maintaining the image details, specific areas of the image are denoised, and thresholds and filter parameters are set to ensure that important details are not affected. Apply adaptive filters, such as contrast-limited adaptive histogram equalization or adaptive gamma correction algorithms, to correct the local contrast of the locally denoised image and enhance the details and layering of the image. Use an image correction algorithm to perform geometric correction on the corrected image to correct perspective distortion or lens distortion to obtain the final stitched image.

[0048] Optionally, the embodiment of the present application performs geometric correction on the corrected image, including:

[0049] The artifacts of the splicing are analyzed in a local area, and the artifacts are identified and eliminated by using a frequency domain transformation; and a quality inspection is performed on an image result after the artifacts are eliminated to obtain the spliced ​​image.

[0050] Specifically, the corrected image is geometrically corrected to identify and eliminate artifacts that may occur during the stitching process to improve the quality of the stitched image. The local area of ​​the stitched image is analyzed, the pixel value differences of adjacent areas are compared, the areas that do not meet the normal image characteristics are identified, and possible artifacts are detected. Among them, artifacts are visual discontinuities or errors that may occur during the image stitching process, usually caused by image registration errors, noise or algorithm limitations. The detected artifact area is converted from the spatial domain to the frequency domain, and the characteristics of the frequency domain transformation are used to identify and eliminate artifacts. For example, the filter can operate on a specific frequency range, and the low-frequency components are removed by applying a high-pass filter, retaining high-frequency details, thereby reducing the impact of artifacts. Then, the processed image is quality tested, and a combination of subjective and objective evaluations is used to ensure that the artifacts have been effectively eliminated and the image quality meets the preset standards. Among them, the subjective evaluation includes the naturalness and continuity of the image observed by the human eye, and the objective evaluation includes calculating indicators such as the PSNR (peak signal-to-noise ratio) of the image. This process effectively guarantees the geometric relationship and visual consistency of the stitched image.

[0051] Furthermore, the embodiment of the present application obtains an image captured at a fixed angle, including:

[0052] The image acquisition device is extracted according to the target scene requirements; the image acquisition impact of light and weather conditions is calculated, and multiple angles are shot according to the image acquisition impact to obtain the fixed-angle shot image, which is marked with location information.

[0053] Specifically, according to the requirements of the target scene, select the appropriate image acquisition device. For example, for shooting that requires high resolution, large format and professional control, a DSLR camera or a medium format camera may be selected; for shooting that requires fast movement and flexible viewing angles, a drone or a sports camera may be selected. Before shooting, analyze the light and weather conditions through professional photography software or weather forecast services, evaluate the impact of light and weather conditions on image acquisition, and determine the best shooting time and angle. According to the evaluation results, shoot from multiple angles, and then select the best image from them, or generate a panoramic or wide-angle image through image stitching technology. During the shooting process, use a GPS device or the built-in GPS function of a mobile phone to record the location information of each image when it is taken, and add location information to each image, including longitude, latitude, altitude, etc. Through these steps, a series of fixed-angle shooting images can be obtained, which not only have consistent angles and location information, but also have guaranteed image quality due to the consideration of the impact of light and weather conditions, providing basic image data for subsequent image stitching and editing.

[0054] In summary, the digital content production method based on image splicing provided in the embodiment of the present application has the following technical effects:

[0055] The image acquisition device is used to acquire images, and the acquired images include images shot at a fixed angle. The overlapping parts of the acquired images reach a preset overlapping threshold, ensuring that there is sufficient consistency and overlapping area between the images, providing a stable basis for subsequent feature point matching, and reducing the stitching error caused by inconsistent shooting angles. The acquired images are subjected to distortion correction to obtain a corrected image, eliminating or reducing the distortion caused by the lens and shooting conditions, improving the authenticity of the image, and reducing the image dislocation caused by distortion during the stitching process. The corrected images are subjected to color adjustment to obtain a corrected color image, eliminating the color difference between different images. The stitched images are made more visually consistent, reducing the visual defects caused by color differences. The scale-invariant feature transformation algorithm is used to extract feature points from the corrected color image, and the acquired images are stitched according to the feature points. After a series of image processing processes, a stitched image is obtained to ensure the quality of the stitched image.

[0056] In general, the embodiments of the present application optimize image acquisition by fixed-angle shooting and preset overlap thresholds, pre-process images through distortion correction and color adjustment, and finally use a scale-invariant feature transformation algorithm to extract and stitch feature points, thereby achieving high-precision and high-efficiency image stitching, and achieving the effect of improving the accuracy of stitched images and the quality of digital content.

[0057] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A digital content production method based on image splicing, characterized in that: include: Capturing images by an image acquisition device, wherein the acquired images include images captured at a fixed angle, and the overlapping parts of the acquired images reach a preset overlapping threshold; Performing distortion correction on the collected image to obtain a corrected image; Performing color adjustment on the corrected image to obtain a corrected color image; A scale-invariant feature transformation algorithm is used to extract feature points from the corrected color image, and the collected images are spliced ​​according to the feature points to obtain a spliced ​​image.

2. The method for producing digital content based on image splicing according to claim 1, characterized in that: The step of stitching the collected images according to the feature points to obtain a stitched image comprises: Compute the computational accuracy of the homography matrix; Optimizing the homography matrix according to the calculation accuracy to obtain an optimized combination; Real-time monitoring of registration errors through the optimization combination, and correction of the registration errors using a feedback loop to obtain a registration result; The registration quality of the registration result is evaluated, and if the registration threshold is met, the homography matrix corresponding to the optimized combination is retained.

3. The method for producing digital content based on image splicing according to claim 1, characterized in that: The step of stitching the collected images according to the feature points to obtain a stitched image includes: Calculating a homography matrix based on the feature points, and obtaining a transformation relationship of the acquired image based on the description of the homography matrix; Performing weighted averaging of the acquired images according to the transformation relationship to obtain a fused image; The visual difference at the image seam of the fused image is eliminated by feathering stitching to obtain the stitched image.

4. The method for producing digital content based on image splicing according to claim 3, characterized in that: Eliminating visual differences at image seams of the fused images by feathering stitching to obtain the stitched image comprises: Algorithm for stitching feature points according to the feature point scene requirements; Adjusting feature point detection parameters according to the feature point algorithm; Calculate the detection impact on the detection result based on the feature point detection parameters; The feature points are initially filtered according to the detection influence, and the erroneous splicing points are removed to generate feature points.

5. The method for producing digital content based on image splicing according to claim 4, characterized in that: Eliminating visual differences at image seams of the fused images by feathering stitching to obtain the stitched image includes: The fused image is cropped to remove the distorted portion of the edge of the image to obtain a cropped image; Performing sharpening processing on the cropped image to obtain a sharpened image; Reducing the noise of the sharpened image by noise removal to obtain a de-noised image; Performing color enhancement on the noise-reduced image according to image requirements to obtain an enhanced image; The enhanced image is format converted to obtain the spliced ​​image.

6. The method for producing digital content based on image splicing according to claim 5, characterized in that: Eliminating visual differences at image seams of the fused image by feathering stitching to obtain the stitched image further includes: Detecting and removing dynamic blur in the stitched image to obtain a deblurred image; Enhance the pixels of the deblurred image using a high dynamic range to obtain a high dynamic range image; Performing local noise reduction on the high dynamic range image according to image content to obtain a local noise reduction image; Using an adaptive filter to correct the local contrast of the local noise reduction image to obtain a corrected image; The corrected image is geometrically corrected to obtain the stitched image.

7. The method for producing digital content based on image splicing according to claim 6, characterized in that: The step of performing geometric correction on the corrected image comprises: Analyze the splicing artifacts by local area, and identify and eliminate the artifacts by frequency domain transformation; The quality of the image result after eliminating artifacts is detected to obtain the stitched image.

8. The method for producing digital content based on image splicing according to claim 1, characterized in that: Get fixed-angle images, including: Extracting the image acquisition device according to the target scene requirements; The influence of light and weather conditions on image acquisition is calculated, and multiple angle shooting is performed according to the influence of image acquisition to obtain the fixed-angle shot image, and the fixed-angle shot image is marked with position information.

Images