CN-122023115-A - Image stitching imaging method suitable for super-resolution reconstruction of large-size high-flux image

CN122023115ACN 122023115 ACN122023115 ACN 122023115ACN-122023115-A

Abstract

The invention discloses an image stitching imaging method and system suitable for super-resolution reconstruction of large-size high-flux images, wherein the method comprises the following steps: and acquiring a large-size high-flux original image, cutting the large-size high-flux original image into a plurality of image sub-blocks based on the hardware performance and reconstruction requirements of the equipment, distributing unique identifiers for the sub-blocks, and associating and recording space position information. And converting the image subblocks into a brightness-chromaticity separation type color space, separating brightness and chromaticity channel images, and storing the chromaticity channel images in a lossless manner. And processing the brightness channel image by adopting a super-resolution reconstruction method with priori constraints to obtain a high-resolution brightness channel sub-block. And splicing the sub-blocks according to the spatial position and the identification information, calling the chroma channel image fusion, eliminating the splice boundary difference, converting back to the original color space, and generating a super-resolution reconstructed spliced image of the original image. The method solves the problems that boundary difference is easy to generate in block splicing, and color distortion is easy to occur in full color reconstruction.

Inventors

GUI DAN
YAN YUXIANG

Assignees

江汉大学

Dates

Publication Date: 20260512
Application Date: 20251223

Claims (10)

1. An image stitching imaging method suitable for super-resolution reconstruction of a large-size high-throughput image, which is characterized by comprising the following steps: S1, acquiring a large-size high-flux original image, and cutting the original image into a plurality of image sub-blocks conforming to reconstruction conditions based on hardware performance threshold values and super-resolution reconstruction adaptation requirements of image processing equipment; S2, converting each obtained image sub-block from an original color space to a brightness-chromaticity separation type color space, separating to obtain a brightness channel image bearing the detail characteristics of the image and a chromaticity channel image bearing the color characteristics of the image, and carrying out lossless storage on the chromaticity channel image to completely retain original color information; s3, performing super-resolution reconstruction on each brightness channel image obtained by separation by adopting a super-resolution reconstruction method with priori constraint, so as to obtain high-resolution brightness channel sub-blocks; S4, according to the recorded spatial position information and the unique identification information, all the obtained high-resolution luminance channel sub-blocks are spliced according to the spatial topological relation of the original image to form a complete high-resolution luminance channel image, the stored chrominance channel image is called to be subjected to color fusion processing with the complete high-resolution luminance channel image, boundary differences generated by sub-block splicing are eliminated synchronously, and then the super-resolution reconstructed spliced image corresponding to the large-size high-flux original image is obtained after the boundary differences are converted back to the original color space.
2. The image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images according to claim 1, wherein the pixel size of the original image in S1 is not less than ten thousand times ten thousand, the data size is not less than 10G, the number of channels is multiple channels, and the bit depth is not less than 8-bit.
3. The image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images according to claim 1, wherein the spatial location information in S1 is determined based on a coordinate system established by a preset reference point of an original image, and the specific process is as follows: Selecting definitely calibrated pixel points on an original image as preset reference points To Establishing a horizontal edge direction of an original image as a coordinate origin Axis, build-up along vertical edge direction A shaft, constructing a two-dimensional rectangular coordinate system, wherein, The positive axis direction is consistent with the horizontal pixel arrangement direction of the original image, The positive direction of the axis is consistent with the arrangement direction of the vertical pixels of the original image, and the coordinate units are pixels; Let the total number of horizontal pixels of the original image be The total number of vertical pixels is Presetting a datum point The pixel coordinates in the original image are , The range of the values is as follows , The range of the values is as follows ; Let the number of horizontal pixels of each image sub-block be The number of vertical pixels is Each sub-block after cutting is orderly distributed according to a rule of row priority or column priority, and for any image sub-block The pixel coordinate of the top left corner vertex in the two-dimensional rectangular coordinate system is The pixel coordinates of the vertex of the lower right corner are Wherein it is satisfied that 、 And (2) and And associating the unique identification information of each image sub-block with the corresponding Storing coordinate information in one-to-one association, wherein the coordinate information directly maps the space occupation range of the sub-block in the original image; wherein the variables are defined as follows: For the total number of pixels contained in the horizontal direction of the original image, For the total number of pixels contained in the vertical direction of the original image, Is a preset datum point At the position of The pixel coordinate values on the axis are, Is a preset datum point At the position of The pixel coordinate values on the axis are, For the number of pixels contained in the horizontal direction of a single image sub-block, For the number of pixels contained in the vertical direction of a single image sub-block, For image sub-blocks The pixel coordinate value of the upper left corner vertex on the X-axis, For image sub-blocks Pixels with the top left corner vertex on the Y-axis, For image sub-blocks The right lower corner is at the vertex The pixel coordinate values on the axis are, For image sub-blocks The right lower corner is at the vertex Pixel coordinate values on the axis.
4. The image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images according to claim 1, wherein the luminance-chrominance separation type color space in S2 includes Ycbcr color space, yuv color space, YIQ color space or other color spaces with luminance and chrominance decoupling function.
5. The image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images according to claim 1, wherein the super-resolution reconstruction method with prior constraint in S3 is characterized in that at least one of sparsity prior and continuity prior is introduced to construct an optimized objective function, and image detail recovery is achieved through iterative solution; The sparsity prior is realized by an L1 norm, an L0 norm or other sparse regularization modes, and the continuity prior is realized by Hessian matrix regularization terms, gradient constraint, total variation constraint or other spatial continuity regularization modes.
6. The image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images according to claim 5, wherein the optimization objective function specifically comprises: , Wherein, the Representing a luminance channel image to be reconstructed; corresponding to the brightness channel image separated after the color space conversion and preprocessed; background components in the luminance channel image; characterizing a point spread function, PSF, of an imaging system; is the square of the L2 norm, made up of As a fidelity term, the core function is to restrict the difference between the reconstructed image and the input brightness channel image; For sparsity prior terms, the sparsity prior terms are realized by combining the distribution characteristics of targets and backgrounds in the images through L1 norms, L0 norms or other sparse regularization modes, and specifically can adopt 、 Or other equivalent sparse constraint forms; Based on the continuity characteristics of the target spatial distribution in the image, the method is realized by using Hessian matrix regularization terms, gradient constraint, total variation constraint or other spatial continuity regularization modes as continuity prior terms Based on different imaging quality and target characteristics of the image, the optimization objective function adjusts the prior term configuration under different conditions in such a way that when the details of the image are fuzzy and the important strengthening details are required to be extracted, the optimization objective function is as follows ; When the image noise is more and the artifact needs to be suppressed, the optimization objective function is that 。
7. The image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images according to claim 1, wherein the process of stitching all the obtained high-resolution luminance channel sub-blocks according to the spatial topological relation of the original image in S4 is as follows: Firstly, unique identification information corresponding to each high-resolution brightness channel sub-block and associated stored space position coordinate information are called, wherein the space position coordinate information comprises the top left corner vertex coordinates of each sub-block in a preset two-dimensional rectangular coordinate system With lower right corner vertex coordinates Determining the relative position relation of each sub-block in the original image space topological structure based on the coordinate information, and determining the row-column distribution sequence among the sub-blocks; Then, constructing a blank spliced canvas consistent with the pixel size of the original image, wherein the total number of horizontal pixels and the total number of vertical pixels of the canvas are respectively consistent with the total number of horizontal pixels of the original image Total number of vertical pixels Keeping consistency, and keeping a coordinate system of the canvas and a coordinate system of the space position coordinates of the sub-blocks in a unified way; According to the relative position relation of each sub-block, mapping each high-resolution brightness channel sub-block to the corresponding position of the blank splicing canvas one by one, and enabling the vertex coordinates of the upper left corner of the sub-block in the mapping process Vertex coordinates of lower right corner The coordinate of the target region on the canvas is completely matched with that of the target region on the canvas, so that the space occupation of the sub-blocks in the canvas is consistent with that in the original image; For the adjacent sub-blocks without overlapping, the adjacent sub-blocks are directly butted according to coordinates, so that the space continuity between the sub-blocks is ensured; after the mapping and the butt joint of all the sub-blocks are completed, a complete high-resolution brightness channel image is formed, the spatial topological structure of the image is completely consistent with that of the original image, and the detail information of each sub-block after super-resolution reconstruction is reserved.
8. The image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images according to claim 7, wherein the method for eliminating boundary differences generated by subblock stitching in S4 is as follows: based on the spatial position coordinate information of each high-resolution brightness channel sub-block, two adjacent sub-blocks are identified And (3) with Setting the horizontal coordinate range of the boundary region in a preset two-dimensional rectangular coordinate system as The vertical coordinate range is The coordinates of any pixel point in the boundary area are Wherein 、 ; Defining sub-blocks At the pixel point The gray value at is Sub-block At the pixel point The gray value at is The gray value of the pixel point after splicing is ; For the overlapping boundary area of adjacent sub-blocks, adopting a gray value average value fusion mode, namely Making the gray transition of the overlapped area continuous; For adjacent boundary areas which are not overlapped but have grey abrupt change, constructing a grey transition model by taking the boundary line as a reference, and taking the horizontal or vertical reference coordinates of the boundary line as Or (b) When the transition is made in the horizontal direction, When the transition is made in the vertical direction, The gray values at two sides of the boundary are linearly graded; In the color fusion stage, sub-blocks are arranged The corresponding chromaticity channel is arranged at the pixel point The chromaticity value at is Sub-block The corresponding chromaticity channel is arranged at the pixel point The chromaticity value at is The chroma value of the fused pixel point is By means of calculations consistent with grey value transitions, i.e. overlapping areas The non-overlapped gradual change area is calculated according to a linear gradual change formula in a corresponding direction, so that the brightness and chromaticity transition are ensured to be consistent in a coordinated manner; Wherein, the Representing any two adjacent high resolution luminance channel sub-blocks, 、 Is the coordinate range of the boundary region, is derived from the spatial positions of adjacent sub-blocks, For the original gray values of adjacent sub-blocks in the boundary region, For the original chrominance values corresponding to the neighboring sub-blocks, And (3) realizing natural transition of the boundary region for the target gray value and the chromaticity value after the boundary region is fused by the mathematical model, and eliminating boundary difference generated by splicing.
9. An image stitching imaging method suitable for super-resolution reconstruction of a large-size high-throughput image, which is characterized by comprising the following steps: The image segmentation and information association unit is used for acquiring a large-size high-flux original image, and cutting the original image into a plurality of image sub-blocks conforming to reconstruction conditions based on a hardware performance threshold value and a super-resolution reconstruction adaptation requirement of the image processing equipment; the color space conversion and chromaticity storage unit is used for converting each obtained image sub-block from an original color space to a brightness-chromaticity separation type color space, separating to obtain a brightness channel image bearing the detail characteristics of the image and a chromaticity channel image bearing the color characteristics of the image, and carrying out lossless storage on the chromaticity channel image to completely retain the original color information; The brightness channel super-resolution reconstruction unit is used for respectively carrying out super-resolution reconstruction on each brightness channel image obtained by separation by adopting a super-resolution reconstruction method with priori constraint to obtain high-resolution brightness channel sub-blocks; The sub-block splicing fusion image generation unit is used for splicing all the obtained sub-blocks of the high-resolution brightness channels according to the recorded spatial position information and unique identification information to form a complete high-resolution brightness channel image, calling the stored chromaticity channel image, carrying out color fusion processing on the chromaticity channel image and the complete high-resolution brightness channel image, synchronously eliminating boundary difference generated by sub-block splicing, and converting the boundary difference back to an original color space to obtain a super-resolution reconstruction spliced image corresponding to the large-size high-flux original image.
10. A computer readable storage medium having stored thereon a computer program, the computer program being executable by a processor to perform an image stitching imaging method as claimed in any one of claims 1-8, suitable for super-resolution reconstruction of large-size high-throughput images.

Description

Image stitching imaging method suitable for super-resolution reconstruction of large-size high-flux image Technical Field The invention belongs to the technical field of image data reconstruction, and particularly relates to an image stitching imaging method and system suitable for super-resolution reconstruction of large-size high-flux images. Background Under the background of rapid development of image processing technology, the application requirements of large-size high-flux images are continuously increased, and the images can bear abundant detail characteristics and spatial distribution information, and play a key role in the fields of medical image analysis, industrial detection, security monitoring, digital city modeling and the like. However, due to factors such as hardware performance of the imaging device, interference of environmental transmission, and running stability of the shooting platform, the obtained original image often has a problem of insufficient resolution, and it is difficult to meet the actual requirement of high-precision analysis. The super-resolution reconstruction technology is a core means for improving the image resolution, but the data volume of a large-size high-flux image is huge, and the direct super-resolution reconstruction can exceed the hardware bearing capacity of conventional image processing equipment, so that the processing efficiency is low, and even the reconstruction process cannot be completed. In the prior art, part of schemes adopt a mode of image blocking processing to relieve hardware pressure, but an effective spatial position correlation mechanism is lacked after blocking, boundary difference easily occurs in the splicing process, and the integral integrity of an image is affected. Meanwhile, most super-resolution reconstruction methods directly process full-color images, the perception difference of human eyes on brightness and chromaticity is not fully utilized, color distortion is easy to occur in the reconstruction process, the visual effect of the images is reduced, and the accuracy requirements of subsequent target recognition and analysis are difficult to meet. The existence of the problems severely restricts the application value of the large-size high-flux image, and how to realize high-resolution reconstruction and high-quality splicing of the image on the premise of adapting to the hardware performance becomes a technical problem to be solved in the current image processing field. The development of the efficient image super-resolution reconstruction and splicing method has important practical significance for improving the application level of images and promoting the technical upgrading of related industries. Disclosure of Invention The invention aims to solve the problems that the large-size high-flux image is difficult to directly reconstruct in super-resolution due to large data volume, boundary difference is easy to generate in block splicing, and color distortion is easy to occur in full-color reconstruction, and provides a super-resolution reconstruction splicing imaging method which is adaptive to hardware performance, realizes independent processing of brightness and chromaticity, eliminates splicing boundary difference, and improves image application value. In view of the above-mentioned drawbacks or improvements of the prior art, as a first aspect of the present invention, the present invention provides an image stitching imaging method suitable for super-resolution reconstruction of large-size high-throughput images, comprising: S1, acquiring a large-size high-flux original image, and cutting the original image into a plurality of image sub-blocks conforming to reconstruction conditions based on hardware performance threshold values and super-resolution reconstruction adaptation requirements of image processing equipment; S2, converting each obtained image sub-block from an original color space to a brightness-chromaticity separation type color space, separating to obtain a brightness channel image bearing the detail characteristics of the image and a chromaticity channel image bearing the color characteristics of the image, and carrying out lossless storage on the chromaticity channel image to completely retain original color information; s3, performing super-resolution reconstruction on each brightness channel image obtained by separation by adopting a super-resolution reconstruction method with priori constraint, so as to obtain high-resolution brightness channel sub-blocks; S4, according to the recorded spatial position information and the unique identification information, all the obtained high-resolution luminance channel sub-blocks are spliced according to the spatial topological relation of the original image to form a complete high-resolution luminance channel image, the stored chrominance channel image is called to be subjected to color fusion processing with the complete high-resolution luminance channel image, bounda