A subsampled scan imaging local motion artifact detection and removal method

Abstract

The present application relates to the technical field of image processing, and belongs to a kind of sub-sampling scanning imaging local motion artifact detection and removal method, the present application is acquired to multiple sub-sampling images under the same field of view: it can automatically detect the structural artifact region introduced due to local motion in scanning process;In the detected artifact region, a reference reconstruction image is introduced for adaptive replacement or fusion;In non-artifact region, try to keep the real detail information in the original sub-sampling image;Thus, under the premise that the real structure resolution of image is not significantly reduced, scanning artifacts, stripe noise or periodic structure noise are effectively suppressed, and the imaging quality and stability of the final fusion image are improved.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, specifically to a method for detecting and removing local motion artifacts in subsampled scanning imaging. Background Technology

[0002] In the fields of microscopic imaging, scanning imaging, and computational imaging, multi-subsampling imaging is often used to overcome the limitations of sensor pixel size or optical system sampling. This involves acquiring multiple sub-sampled images with sub-pixel offsets within the same field of view by changing the sampling phase, sampling position, or imaging path, and then fusing these sub-sampled images to improve the effective sampling rate or spatial resolution of the final image. For example, in scanning light field microscopy systems, it is often necessary to acquire multiple sub-sampled image data within the same field of view and then fuse these sub-sampled images to obtain a high-resolution or high-sampling-rate image. However, in the aforementioned multi-subsampling scanning imaging process, due to the rapid local cell movement speed of biological samples, local periodic stripes, artifact textures, or structural noise can easily appear in the reconstruction results. These artifacts typically exhibit a regular or semi-regular spatial distribution, thus affecting the visual quality and quantitative analysis accuracy of the final reconstructed image.

[0003] In existing technologies, common processing methods include global filtering, simple threshold denoising, or directly replacing the original image with the reconstructed result. However, these methods either tend to over-smooth realistic details or fail to differentiate between real structures and scanning artifacts, thus introducing detail loss or structural distortion while suppressing artifacts. Therefore, there is an urgent need for an image processing method that can effectively detect and suppress scanning artifacts or structural noise in multi-subsampling scanning imaging scenarios while preserving realistic image details as much as possible. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for detecting and removing local motion artifacts in subsampled scanning imaging. This invention can effectively distinguish between real structures and scanning artifacts or structural noise in multi-subsampled scanning imaging scenarios. By introducing a reference reconstructed image for adaptive fusion processing, it significantly improves the stability and visual quality of the reconstructed image while ensuring that image details are preserved as faithfully as possible.

[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a method for detecting and removing local motion artifacts in subsampled scanning imaging, comprising the following steps: S1. Within the same field of view, multiple sub-sampled images with different sampling positions are acquired by a scanning imaging system, and there is a sub-pixel-level sampling offset relationship between the sub-sampled images; S2. Based on the subpixel-level offset relationship between subsampled images, perform interleaved pixel fusion on the subsampled images to generate a directly fused image; S3. Based on the multi-subsampled image, a reconstruction algorithm is used to generate a reference reconstructed image; S4. Perform feature response calculation on the directly fused image to obtain the artifact feature response map of the directly fused image; S5. Based on the artifact feature response map, through threshold segmentation and smoothing, generate an artifact region mask image to indicate the regions in the subsampled image that may be affected by scanning artifacts or structural noise. S6. Based on the artifact region mask image, perform adaptive fusion processing on the reference reconstructed image and the fused image to obtain an artifact-free image.

[0006] Preferably, the feature response calculation in step S4 specifically includes: The directly fused image is convolved using a two-dimensional convolution kernel, and the convolution result is normalized to obtain the artifact feature response of the directly fused image. The structure of the two-dimensional convolution kernel is determined by the scanning sampling period.

[0007] Preferably, the threshold segmentation in step S5 specifically includes: suppressing the regions in the artifact feature response map where the pixel grayscale value is lower than the threshold value to the background region and setting the pixel value to 0, thereby obtaining the segmentation result map.

[0008] Preferably, the smoothing process in step S5 specifically includes: performing Gaussian smoothing on the segmentation result image to obtain a mask image of the artifact region.

[0009] Preferably, the adaptive fusion processing in step S6 specifically includes: using the pixel values ​​of the reference reconstructed image at the location indicated by the artifact region mask; and using the pixel values ​​of the directly fused image at the location indicated by the artifact region mask as a non-artifact region.

[0010] Preferably, the method further includes: S7. Perform steps S1-S6 on the subsampled images of multiple channels respectively, and output the corresponding artifact suppression results, or further perform multi-channel fusion to obtain the final image; Each of the multiple channels satisfies the following condition: multiple subsampled images acquired within that channel correspond to the same field of view.

[0011] Compared with the prior art, the present invention provides a method for detecting and removing local motion artifacts in subsampling scanning imaging, which has the following beneficial effects: This invention can effectively distinguish between real structures and scanning artifacts or structural noise in multi-subsampled scanning imaging scenarios. By introducing a reference reconstructed image for adaptive fusion processing, it significantly improves the stability and visual quality of the reconstructed image while ensuring that image details are preserved as much as possible.

[0012] The features and advantages of the present invention will be described in detail through embodiments and in conjunction with the accompanying drawings. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the steps of a subsampling scanning imaging local motion artifact detection and removal method according to the present invention. Figure 2 The images shown are obtained from embodiments of the present invention, where Figure a is a directly fused image, Figure b is a reference reconstructed image, Figure c is an artifact mask image, and Figure d is an artifact-removed image. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. However, it should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0015] See Figures 1-2 In this embodiment, the invention will be described using subsampled image data acquired by a multi-subsampled scanning imaging system as an example.

[0016] Step 1: Within the same biological sample field of view (such as the field of view for live cell imaging), use a scanning light field microscope system 3 3-subpixel scanning acquires multiple subsampled images, including low-resolution subsampled images. The size is Where H and W are the height and width of a single subsampled image. The 9 subsampled images correspond to 3 A 3-pixel subpixel scan grid with an offset step of 1 / 3 pixel between adjacent scan positions.

[0017] Step 2: Extract the 9 subsampled images Pixels are rearranged in an interleaved manner according to the scan position to achieve direct fusion of subsampled images. For the subimage corresponding to the i-th row and j-th column in the scan grid, its pixels are filled into the directly fused image according to the following mapping relationship. Directly fused images Size is 3H 3W : ; That is, the pixels of each sub-image are filled into the target image at periodic positions with a row and column step size of 3, so that the 9 sub-images are spatially staggered, thereby forming a high-resolution fused image with a size of 3H×3W.

[0018] Step 3: Extract the 9 sub-sampled images Input a pre-trained temporal image super-resolution model to obtain a reference reconstructed image. The dimensions are 3H×3W.

[0019] Step 4: Directly merged images Perform artifact feature response calculation.

[0020] Specifically, a two-dimensional convolution kernel K is used for the directly fused image. Perform convolution processing and normalize the convolution result to obtain the artifact feature response R of the directly fused image: The convolution kernel K has a structure related to the scanning sampling period, used to enhance the response to periodic fringes or structural artifacts generated during the scanning process, wherein... .

[0021] Step 5: Generate an artifact region mask image M based on the artifact feature response R of the directly fused image.

[0022] Specifically, the feature response map can first be thresholded to suppress regions with a value below 0.4 as background regions, with pixel values ​​set to 0. Then, the result after thresholding is Gaussian smoothed to enhance the continuity of the regions, thereby obtaining the artifact region mask image. M>0 indicates the detected artifact region, M=0 indicates the normal region, and M is the pixel value of the artifact region mask image.

[0023] Step 6: Directly fuse the images based on the mask. Reconstructed image with reference Adaptive fusion processing is performed to obtain the artifact-free image I. Specifically, in the locations where the artifact region mask indicates the artifact region, the pixel values ​​of the reference reconstructed image are used; in the non-artifact regions, the pixel values ​​of the directly fused image are used. The method described in this embodiment can effectively detect and suppress periodic stripe artifacts or structural noise introduced by the scanning or reconstruction process in multi-subsampling scanning imaging scenarios, while preserving the original image details in non-artifact areas, thereby improving the stability and visual quality of the final imaging results.

[0024] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions or improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting and removing local motion artifacts in subsampling scanning imaging, characterized in that: Includes the following steps: S1. Within the same field of view, multiple sub-sampled images with different sampling positions are acquired by a scanning imaging system, and there is a sub-pixel-level sampling offset relationship between the sub-sampled images; S2. Based on the subpixel-level offset relationship between the subsampled images, perform interleaved pixel fusion on multiple subsampled images to generate a directly fused image; S3. Based on the multi-subsampled image, a reconstruction algorithm is used to generate a reference reconstructed image; S4. Perform feature response calculation on the directly fused image to obtain the artifact feature response map of the directly fused image; S5. Based on the artifact feature response map, through threshold segmentation and smoothing, generate an artifact region mask image to indicate the regions in the subsampled image that may be affected by scanning artifacts or structural noise. S6. Based on the artifact region mask image, perform adaptive fusion processing on the reference reconstructed image and the fused image to obtain an artifact-free image.

2. The method for detecting and removing local motion artifacts in subsampling scanning imaging according to claim 1, characterized in that: The generation of the direct fused image in step S2 specifically includes: For a subsampled image corresponding to a certain position in the scan grid, its pixels are filled into the directly fused image according to the mapping relationship from low-resolution image to high-resolution image.

3. The method for detecting and removing local motion artifacts in subsampling scanning imaging according to claim 1, characterized in that: The feature response calculation in step S4 specifically includes: The directly fused image is convolved using a two-dimensional convolution kernel, and the convolution result is normalized to obtain the artifact feature response of the directly fused image. The structure of the two-dimensional convolution kernel is determined by the scanning sampling period.

4. The method for detecting and removing local motion artifacts in subsampling scanning imaging according to claim 1, characterized in that: The threshold segmentation in step S5 specifically includes: suppressing regions in the artifact feature response map with pixel gray values ​​lower than a threshold value to background regions and setting the pixel values ​​to 0, thereby obtaining a segmentation result map.

5. The method for detecting and removing local motion artifacts in subsampling scanning imaging according to claim 4, characterized in that: The smoothing process in step S5 specifically includes: performing Gaussian smoothing on the segmentation result image to obtain a mask image of the artifact region.

6. The method for detecting and removing local motion artifacts in subsampling scanning imaging according to claim 1, characterized in that: The adaptive fusion process in step S6 specifically includes: using the pixel values ​​of the reference reconstructed image at the location indicated by the artifact region mask; and using the pixel values ​​of the directly fused image at the location indicated by the artifact region mask as a non-artifact region.

7. The method for detecting and removing local motion artifacts in subsampling scanning imaging according to claim 1, characterized in that: The method further includes: S7. Perform steps S1-S6 on the subsampled images of multiple channels respectively, and output the corresponding artifact suppression results, or further perform multi-channel fusion to obtain the final image; Each of the multiple channels satisfies the following condition: multiple subsampled images acquired within that channel correspond to the same field of view.

Images