CN-121981899-A - Image fusion method, device, equipment and medium based on improved optimal suture line

CN121981899ACN 121981899 ACN121981899 ACN 121981899ACN-121981899-A

Abstract

The application discloses an image fusion method, device, equipment and medium based on improved optimal suture line, belonging to the technical field of image processing, wherein the method comprises the steps of acquiring two preprocessed images to be registered, wherein the images have overlapping areas; the method comprises the steps of constructing an energy function based on color difference, structure difference, centroid difference and texture information of pixels in an overlapping area, determining composite energy values of pixel points in the overlapping area according to the function, searching and determining a path with the smallest accumulated energy value in the overlapping area to serve as an optimal suture line based on the composite energy values of all the pixel points by adopting a global path optimization algorithm, and carrying out pixel fusion on two images to be registered along the optimal suture line, wherein the pixel fusion is carried out in a fusion area surrounding the optimal suture line, and the width of the fusion area is adjusted according to the texture information in the fusion area.

Inventors

XIANG PEI
SHI JIN
SUN RUIYANG
SUN YUNHAN
WANG BINGJIAN
LAI RUI
QIN HANLIN
GAO MENGYANG
ZHOU HUIXIN
YANG TIANFANG
QI SHUXIA
SONG JIANGLUQI
LI CHONG
Xu Xiujin
PU ZHENG

Assignees

西安电子科技大学
中国石油大学(北京)克拉玛依校区

Dates

Publication Date: 20260505
Application Date: 20251230

Claims (10)

1. An improved optimal suture-based image fusion method, comprising: acquiring two preprocessed images to be registered, wherein the two images to be registered have an overlapping area; Constructing an energy function based on color difference, structure difference, centroid difference and texture information of pixels in an overlapping region, and determining a composite energy value of pixel points in the overlapping region according to the energy function; Searching and determining a path with the smallest accumulated energy value in the overlapped area by adopting a global path optimization algorithm based on the composite energy values of all pixel points, and taking the path as an optimal suture line; and carrying out pixel fusion on the two images to be registered along the optimal suture line, wherein the pixel fusion is carried out in a fusion area surrounding the optimal suture line, and the width of the fusion area is adjusted according to texture information in the fusion area.
2. The improved optimal stitch line-based image fusion method as recited in claim 1, wherein the constructing an energy function based on color differences, structural differences, centroid differences, and texture information, comprises: Converting the image to be registered from an RGB color space to an HSL color space, and calculating the color difference based on a hue difference, a saturation difference and a brightness difference; Determining the structural difference in combination with a Scharr operator and a modified Sobel operator, wherein the modified Sobel operator comprises a +45° direction sum A convolution kernel at 45 ° direction; determining the centroid difference through Euclidean distance of gray centroids in a preset neighborhood range with corresponding pixel positions in the two images to be registered as centers; extracting texture information of the image to be registered by using a Gabor filter bank, and processing a texture information map by using gray morphology closing operation to obtain texture differences; And carrying out weighted summation on the color difference, the structure difference, the regional centroid difference and the texture difference to obtain the energy function.
3. The improved optimal stitch line based image fusion method as recited in claim 2, wherein the color difference of the overlapping area pixels is determined according to the following formula: Wherein Δh represents the hue difference of the pixels corresponding to the two images to be registered, S 1 (x, y) and S 2 (x, y) represent the saturation values of the pixels corresponding to the two images to be registered, and L 1 (x, y) and L 2 (x, y) represent the brightness values of the pixels corresponding to the two images to be registered.
4. The improved optimal seam line based image fusion method of claim 2, wherein the structural differences of the overlapping area pixels are determined according to the following formula: Wherein S +45° and S 45° Respectively the sum of the Sobel operator in the +45 DEG direction The convolution kernels in the 45-degree direction, I 1 (x, y) and I 2 (x, y) are pixel gray values, E ag (x, y) is an improved gradient difference in the oblique direction calculated by the Sobel operator, E bg (x, y) is a gradient difference in the horizontal and vertical directions calculated by the Scharr operator, and S chx and S chy are convolution kernels in the x-direction and the y-direction of the Scharr operator, respectively.
5. The improved optimal suture-based image fusion method of claim 2, wherein the energy function is determined according to the following formula: Wherein χ is a preset adjustment factor, E newc (x, y) is the color difference of the overlapping area pixels, E newg (x, y) is the structure difference of the overlapping area pixels, E c (x, y) is the centroid difference of the overlapping area pixels, and E gabor (x, y) is the texture information of the overlapping area pixels.
6. The improved optimal seam based image fusion method of claim 1, wherein the composite energy value, using a global path optimization algorithm, searches for and determines a path with the smallest cumulative energy value in the overlapping region as the optimal seam, comprising: Creating a pixel stitching direction recording matrix with the same size as the overlapping area and an accumulated energy value array with the same width as the overlapping area, wherein the array is used for recording accumulated energy values of stitching lines at different starting positions, and the initial value is the energy value of the first row of pixels of the overlapping area; Starting from the second row of the overlapping area, for each pixel point of the current row, searching a point with the smallest accumulated energy value in a plurality of adjacent pixel points and possible characteristic points of the previous row, updating the accumulated energy value array by adding the accumulated energy value of the current pixel point, recording a stitching direction to the stitching direction recording matrix, stopping when the last row of pixels is reached, searching the minimum value in the accumulated energy value array, starting from the pixel position corresponding to the minimum value, tracing back a searching path upwards according to the stitching direction recording matrix, and determining the optimal stitching line.
7. The improved optimal seam-based image fusion method of claim 1, wherein the pixel fusion of the two images to be registered along the optimal seam, wherein the pixel fusion is performed in a fusion area surrounding the optimal seam, and the width of the fusion area is adjusted according to texture information in the fusion area, comprising: The method comprises the steps of presetting a basic fusion width as w, taking an optimal suture line as a central line, extending w pixels to the left side and the right side to form a fusion area, and carrying out gradual-in gradual-out fusion in the fusion area according to the following formula: Wherein S e is the abscissa of the pixel on the stitching line, 2w is the preset width value of the gradually-in and gradually-out fusion area, and if the image texture exists in the w/2 range on both sides of the optimal stitching line, the fusion width of the local area is adjusted to w/2.
8. An image fusion apparatus based on an improved optimal suture thread, the apparatus comprising: the system comprises an acquisition unit, a function construction unit, a function generation unit, a control unit and a control unit, wherein the acquisition unit is used for acquiring two preprocessed images to be registered, the two images to be registered have an overlapping area; a suture line determining unit, configured to search and determine a path with the smallest cumulative energy value in the overlapping area as an optimal suture line by using a global path optimization algorithm based on the composite energy values of all the pixel points; and the pixel fusion unit is used for carrying out pixel fusion on the two images to be registered along the optimal suture line, wherein the pixel fusion is carried out in a fusion area surrounding the optimal suture line, and the width of the fusion area is adjusted according to texture information in the fusion area.
9. A computer device comprising a memory and a processor, the memory storing a computer program that, when executed by the processor, causes the processor to perform the steps of the method of any of claims 1 to 7.
10. A computer readable storage medium, characterized in that a computer program is stored, which, when being executed by a processor, causes the processor to perform the steps of the method according to any of claims 1 to 7.

Description

Image fusion method, device, equipment and medium based on improved optimal suture line Technical Field The invention belongs to the technical field of image processing, and particularly relates to an image fusion method, device, equipment and medium based on improved optimal suture lines. Background Among the existing image fusion techniques, the optimal suture-based method has a good visual effect. However, the method has obvious defects that the energy function is mostly different in RGB color space calculation, the matching degree with human visual perception characteristics is low, the structural difference calculation usually depends on a Sobel operator in a single direction, the perception of geometric distribution characteristics of a pixel neighborhood is lacked, the influence of a region penetrated by a suture line on the final fusion quality is not judged, the found suture line cannot be optimal, the gray offset information of a low texture region is not perceived sufficiently, a texture feature extraction mechanism is not fused effectively, the suture line is easy to pass through a region with poor alignment, structural dislocation or content deletion of a fusion result is caused, the global searching capacity of a dynamic programming algorithm is limited, a local optimal path is easy to fall into, the surrounding processing range of the suture line is fixed, the residual obvious suture trace is caused due to the too small range, the visual blurring is caused due to the too large range, and the relation between the fusion trace elimination and the image definition is difficult to balance. Disclosure of Invention Based on the above, it is necessary to provide an image fusion method, device, equipment and medium based on improved optimal suture line, aiming at solving the problems of easy structural dislocation, obvious suture trace and local blurring of spliced images caused by unreasonable energy function design and stiff fusion mode in the prior art. The embodiment of the application provides an image fusion method based on an improved optimal suture, which comprises the following steps: acquiring two preprocessed images to be registered, wherein the two images to be registered have an overlapping area; Constructing an energy function based on color difference, structure difference, centroid difference and texture information of pixels in an overlapping region, and determining a composite energy value of pixel points in the overlapping region according to the energy function; Searching and determining a path with the smallest accumulated energy value in the overlapped area by adopting a global path optimization algorithm based on the composite energy values of all pixel points, and taking the path as an optimal suture line; and carrying out pixel fusion on the two images to be registered along the optimal suture line, wherein the pixel fusion is carried out in a fusion area surrounding the optimal suture line, and the width of the fusion area is adjusted according to texture information in the fusion area. Further, the constructing an energy function based on the color difference, the structure difference, the centroid difference, and the texture information includes: Converting the image to be registered from an RGB color space to an HSL color space, and calculating the color difference based on a hue difference, a saturation difference and a brightness difference; Determining the structural difference in combination with a Scharr operator and a modified Sobel operator, wherein the modified Sobel operator comprises a +45° direction sum A convolution kernel at 45 ° direction; determining the centroid difference through Euclidean distance of gray centroids in a preset neighborhood range with corresponding pixel positions in the two images to be registered as centers; extracting texture information of the image to be registered by using a Gabor filter bank, and processing a texture information map by using gray morphology closing operation to obtain texture differences; And carrying out weighted summation on the color difference, the structure difference, the regional centroid difference and the texture difference to obtain the energy function. Further, the color difference of the overlapping area pixels is determined according to the following formula: Wherein Δh represents the hue difference of the pixels corresponding to the two images to be registered, S 1 (x, y) and S 2 (x, y) represent the saturation values of the pixels corresponding to the two images to be registered, and L 1 (x, y) and L 2 (x, y) represent the brightness values of the pixels corresponding to the two images to be registered. Further, the structural difference of the overlapping region pixels is determined according to the following formula: Wherein S +45° and S 45° Respectively the sum of the Sobel operator in the +45 DEG directionThe convolution kernels in the 45-degree direction, I 1 (x, y) and I 2 (x, y) are pixel gray