Processing method suitable for core image correction and image fusion

Abstract

The invention discloses an image correction and image fusion method suitable for core acquisition, and the method mainly comprises the following steps: carrying out the feature point detection of an acquired original core image with a single overlapping region, and obtaining a longitudinal offset and a transverse offset; the offset factor is calculated, the offset of the line is obtained according to the product of the line number and the offset factor, the offset is adjusted, the operation is repeated until the longitudinal offset is reached, and a rectangle is taken as a corrected image after processing. And performing feature point extraction matching on the corrected image to obtain a to-be-fused region, performing HIS space-based brightness smoothing on a right image in the to-be-fused region to prevent a brightness difference caused by a lamplight irradiation angle difference, and obtaining a processed to-be-fused right image. And according to the obtained to-be-fused image, calculating through an improved energy function to obtain an optimal suture line. And carrying out image fusion through the obtained optimal suture line, and copying the image to a fusion region of the image to obtain a large-viewing-angle rock core image.

Description

Technical Field

[0001] The invention relates to core image correction and image fusion technology, and in particular to a processing method suitable for core image splicing, and belongs to the field of computer vision. Background Art

[0002] Core images are columnar rocks taken from geological layers by researchers using special coring equipment. However, cores may be weathered and other changes may occur under the influence of weather, making it impossible to obtain original geological information. Therefore, cores need to be collected in a timely manner and digitally preserved. To obtain high-resolution complete digital images of cores, multiple core images need to be stitched together. The core surface image is collected by the relative movement of the camera and the core, and multiple core images are collected by the translation of the camera. Therefore, the study of automated image stitching technology can free up manpower and improve stitching efficiency and stitching accuracy.

[0003] Image stitching is to stitch multiple pictures with overlapping parts into a picture with a wide viewing angle. It is generally divided into stitching methods based on frequency domain and stitching methods based on time domain. The stitching method based on frequency domain displays image information in high frequency and low frequency through Fourier transform, solves the transformation vector by calculating the cross power spectrum, and realizes image stitching according to the transformation vector. The stitching method based on time domain is further divided into two categories: stitching based on image feature points and stitching based on grayscale information. The stitching method based on grayscale information is less used in practical applications, and this method is often difficult to achieve good results in practical stitching. The stitching based on image feature points detects the feature points in the image, calculates the relationship between the detected feature points, and realizes image stitching according to the calculated relationship. This method has fast calculation speed and strong robustness, and can achieve good stitching effect. Therefore, this method is used more frequently in practical applications.

[0004] Image fusion is an important step in image stitching. Usually, two images will have visual differences in their overlapping areas because the acquisition conditions are not exactly the same. Selecting a suitable image fusion method can reduce blur, color difference and artifacts caused by factors such as exposure difference and image registration accuracy. Common fusion methods include linear fusion. This method cannot achieve ideal stitching effect when the input images have large differences, and will produce certain ghosting phenomena. The method based on the seam line stitches the images by calculating the best seam line on the overlapping area of ​​the images. The proposed method combining the Laplace pyramid and the best seam line fusion can generally achieve good results. However, this method may result in unnatural brightness transition and obvious seams in the stitching results when the exposure difference of the core images is large. In addition, the rubber roller and the core may move slightly during the core acquisition process, resulting in the offset of the pixels of the acquired image and affecting the quality of subsequent image processing. Summary of the invention

[0005] The technical problem to be solved by the present invention is: to provide a method suitable for core image correction and image fusion, which has solved the problems existing in the above-mentioned prior art;

[0006] The present invention proposes a method for core image correction and image fusion, which mainly includes the following steps:

[0007] (1) Perform feature point detection on the core original image with a single overlapping area to obtain the longitudinal offset and the lateral offset. The offset factor is calculated (offset near-line linear process). The offset of the row is obtained by multiplying the number of rows by the offset factor. The row is adjusted by this offset. Repeat the operation until the longitudinal offset is reached. The processed image can be regarded as a parallelogram, and the rectangle in it is taken as the corrected image.

[0008] (2) Treat all stitched images as described in step (1) to obtain the corrected image;

[0009] (3) According to the image in step (2), the area to be fused is obtained by feature point extraction and matching, and the right image in the area to be fused is brightness smoothed based on the HIS space to prevent the brightness difference caused by the difference in lighting angles, and the processed right image to be fused is obtained;

[0010] (4) According to the right image to be fused and the left image to be fused obtained in step (3), an energy function of HIS information is added to calculate the energy map to obtain the optimal suture line;

[0011] (5) The best stitching line obtained in step (4) is used to fuse the images and is copied to the fusion area of ​​the image to obtain a core image with a wide viewing angle. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a schematic diagram of core collection.

[0013] Figure 2 It is a single core collection map.

[0014] Figure 3 It is a schematic diagram of core correction.

[0015] Figure 4 This is a schematic diagram of brightness processing based on wavelet transform.

[0016] Figure 5 This is a schematic diagram of brightness processing.

[0017] Figure 6 This is the best suture display.

[0018] Figure 7 It is the fusion result and the wide-angle core result map. DETAILED DESCRIPTION

[0019] The present invention will be further described below in conjunction with the accompanying drawings:

[0020] A method suitable for core image correction and image fusion, which can be divided into the following steps:

[0021] (1) Perform feature point detection on the core original image with a single overlapping area to obtain the longitudinal offset and the lateral offset. Calculate the offset factor based on the offset (offset near-line linear process). The offset of the row is obtained by multiplying the number of rows by the offset factor. The row is adjusted by this offset. Repeat the operation until the longitudinal offset is reached. The processed image can be regarded as a parallelogram, and the rectangle in it is taken as the corrected image.

[0022] (2) Treat all stitched images as described in step (1) to obtain the corrected image;

[0023] (3) Based on the image in step (2), the area to be fused is obtained by feature point extraction and matching, and the right image in the area to be fused is brightness smoothed based on the HIS space to prevent brightness differences caused by differences in lighting angles, thereby obtaining the processed right image to be fused;

[0024] (4) According to the right image to be fused and the left image to be fused obtained in step (3), an energy function of HIS information is added to calculate the energy map to obtain the optimal suture line;

[0025] (5) The best stitching line obtained in step (4) is used to fuse the images and is copied to the fusion area of ​​the image to obtain a core image with a wide viewing angle.

[0026] Specifically, Figure 2 As shown, in step (1), the longitudinal offset and the lateral offset are obtained by performing surf extraction on the upper and lower overlapping areas of a single image and filtering the feature points, and the ratio is obtained by the offset, and the pixels are adjusted accordingly as follows:

[0027]

[0028] off l =l×tanθ

[0029] P(x,y)=P(x-off l ,y)

[0030] Among them, offset x and offset y Represents the pixel offset of a single image, the height of the complete core image, and the circumference of the columnar core, respectively. lrepresents the offset of the lth row of the image, × represents the multiplication of the number of rows and the offset scale factor, P(x,y) represents the pixel value, and P(x-offset l ,y) represents the actual pixel value of the corresponding position.

[0031] In step (2), each image is corrected to obtain the corrected image as shown in Figure 3 As shown in FIG. 1 , considering that due to the position of the light source and the core material, the collected images may have large differences in the overlap area, the brightness of the overlap area obtained in step (3) is first processed as follows: Figure 4 As shown, the processed overlapped right image is as follows Figure 5 , where the high frequency sub-band is processed as shown in the following formula:

[0032] h new =w1h left +w2h right

[0033] where h new represents the high-frequency component after processing, h left and h right They represent the high-frequency components of the overlapping left and right images respectively, w1 and w2 represent the corresponding weights respectively, where w1+w2=1.

[0034] The energy diagram of the overlapped right image and the overlapped left image obtained in step (3) is calculated as follows: Figure 6 , where the energy calculation function improves the traditional grayscale component into the HIS component to obtain more robust energy information. The energy function is as follows:

[0035] E(x,y)=w1E geometry (x,y)+w2E hsi (x,y)

[0036]

[0037] Where E(x,y) represents the energy difference of the corresponding pixels, E geometry (x, y) represents the difference in structural information at that point, E hsi (x, y) represents the color information of the point, w1 and w2 represent the weights, where w1+w2=1. 0-l and I 1-l Represents the l component of the left and right images. l includes the three components of HIS, w l Represents the corresponding component weight, and its sum is 1.

[0038] Finally, the best stitching line is obtained through step (5), image fusion is completed, and a fused image is obtained, and finally a core image with a large viewing angle is obtained as shown in FIG. Figure 7At the same time, in order to verify the quality of the image after stitching in this paper, the algorithm of the Peak Signal to Noise Ratio (PSNR) model and the edge difference spectrum evaluation method (Difference of edgemap, DoEM) are selected as evaluation indicators, and compared with the traditional best stitching line. The experimental results are shown in Table 1. It can be seen that compared with the traditional best stitching line, the present invention has improved both indicators.

[0039] Table 1

[0040] method PSNR(dB) Doem Traditional best suture 23.84 1.66 The present invention 27.63 1.91

[0041] In summary, compared with the traditional optimal stitching line, the present invention has obvious advantages in core image fusion. At the same time, an image correction algorithm is proposed for the case of pixel offset in core acquisition, which improves the quality of image processing. Therefore, the present invention is an effective core image correction and image fusion method.

[0042] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for core image correction and image fusion, characterized in that The core image has high resolution, delicate colors and high image quality. Due to acquisition factors, there will be pixel offset and color difference. A fusion method based on pixel correction and wavelet fusion combined with the best stitching line is designed. After pixel correction and brightness fusion of the acquired images, the core image with a good wide viewing angle is finally spliced. It can be divided into the following steps: (1) Perform feature point detection on the core original image with a single overlapping area to obtain the longitudinal offset and the lateral offset. The offset factor is calculated (offset near-line linear process). The offset of the row is obtained by multiplying the number of rows by the offset factor. The row is adjusted by this offset. Repeat the operation until the longitudinal offset is reached. The processed image can be regarded as a parallelogram, and the rectangle in it is taken as the corrected image. (2) Treat all stitched images as described in step (1) to obtain the corrected image; (3) Based on the image in step (2), the area to be fused is obtained by feature point extraction and matching, and the right image in the area to be fused is subjected to brightness fusion based on the HIS space to prevent the brightness difference caused by the difference in lighting angles, thereby obtaining the processed right image to be fused; (4) According to the right image to be fused and the left image to be fused obtained in step (3), an energy function of HIS information is added to calculate the energy map to obtain the optimal suture line; (5) performing image fusion using the best stitching line obtained in step (4) and copying it to the fusion area of ​​the image to obtain a core image with a wide viewing angle; Specifically, in step (1), the vertical offset and horizontal offset are obtained by performing surf extraction on the upper and lower overlapping areas of a single image and filtering the feature points, and the ratio is obtained by the offset, and the pixels are adjusted accordingly as follows: off l =l×tanθ P(x,y)=P(x-off l ,y) Among them, offset x and offset y Represents the pixel offset of a single image, the height of the complete core image, and the circumference of the columnar core, respectively. l represents the offset of the lth row of the image, × represents the multiplication of the number of rows and the offset scale factor, P(x,y) represents the pixel value, and P(x-offset l ,y) represents the actual pixel value of the corresponding position; In step (2), each image is corrected to obtain a corrected image as shown. Considering that the collected images may have large differences in the overlap area due to the position of the light source and the core material, the brightness of the overlap area obtained in step (3) is first processed as shown to obtain the processed overlap right image, where the high frequency sub-band is processed as shown in the following formula: h new =w1h left +w2h right where h new represents the high-frequency component after processing, h left and h right They represent the high-frequency components of the overlapping left and right images, respectively, w1 and w2 represent the corresponding weights, where w1+w2=1; The overlapped right image and the overlapped left image obtained in step (3) are subjected to energy map calculation, wherein the energy calculation function improves the traditional grayscale component into the HIS component to obtain more robust energy information. The energy function is as follows: E(x,y)=w1E geometry (x,y)+w2E hsi (x,y) Where E(x,y) represents the energy difference of the corresponding pixels, E geometry (x, y) represents the difference in structural information at that point, E hsi (x, y) represents the color information of the point, w1 and w2 represent the weights, where w1+w2=1, I 0-l and I 1-l Represents the l component of the left and right images. l includes the three components of HIS, w l represents the corresponding component weight, and its sum is 1; Finally, the best stitching line is obtained through step (5), image fusion is completed, a fused image is obtained, and finally a core image with a wide viewing angle is obtained.

Images