gaussian filter matlab function imgaussfilt  (MathWorks Inc)

Journal: Sensors (Basel, Switzerland)

Article Title: A New Photographic Reproduction Method Based on Feature Fusion and Virtual Combined Histogram Equalization

doi: 10.3390/s21186038

Figure Lengend Snippet: Proposed mutually hybrid histogram analysis approach. We analyze the global and local histograms simultaneously using different statistical methods, i.e., Gaussian mixture model for the former and stratified sampling for the latter. Although different statistical methods are applied, we aim to extract the mutually compatible features to form a virtual combined histogram (introduced later in ).

Article Snippet: A weight map function ( τ i , j ) is generated by convolving the binary map with a Gaussian low-pass filter (the Matlab inbuilt imgaussfilt function) to smooth the weighting difference.

Techniques: Sampling

Journal: Journal of Cell Science

Article Title: CTRL – a label-free artificial intelligence method for dynamic measurement of single-cell volume

doi: 10.1242/jcs.245050

Figure Lengend Snippet: DIC image and cell topography fluorescence image pre-processing. (A) DIC image pre-processing. For each DIC image, we first cropped the image to obtain single cells with the least amount of background (A1). We then acquired a background mask by taking a 10 pixel band closest to the edge (A2). A polynomial fitting using the function ‘poly22’ in MATLAB was applied to perform an intensity fitting to the background mask (A3). The image in A3 is subtracted from the image in A1 to obtain a background-corrected image A4. The purpose of this background correction is to reduce the potential local intensity heterogeneity within one DIC image. After background correction, we acquired a new background mask A5 based on the background-corrected image with the same procedure and created a new artificial background ‘canvas’ for cell padding A6. The ‘canvas’ of size 512×512 was created by generating a Gaussian distribution with the same mean and standard deviation as the background mask acquired in A5. The background-corrected cell in A4 was then put into the middle of A6 to finalize the background padding (A7). Finally, we normalized the image to the data type of uint8 (8-bit) with integer values between 0 and 255 (A8). (B) Cell topography fluorescence image pre-processing. A background intensity polynomial fitting (B2) was first applied to the original fluorescence image (B1) using the same procedure as described in the ‘DIC image pre-processing’ section. The estimated height of the cell at each pixel hpixel is directly proportional to the loss of intensity at the pixel. For instance, if the pixel intensity is 0, then the height at the pixel is the height of the microchamber; if the intensity is the maximum intensity, then the height of the cell at the pixel is 0:hpixel = (1-Ipixel/Ichannel) × hchannel. The relative pixel intensity (B3) Ipixel/Ichannel was then obtained by dividing B1 by B2 for each pixel. The image B3 was then subtracted by 1 and multiplied by the value of microchamber height hchannel to reflect the height of the cell at each pixel (in μm units) as shown in B4. (Note due to optical effects, the intensity only estimates the height. However, the integrated intensity over the image reports the true volume image.) An artificial background ‘canvas’ (B5) was created for padding using the same procedure as described in the ‘DIC image pre-processing’ section. The cell image was put into the center of the ‘canvas’ as shown in B6, and a binary-valued cell mask was created by thresholding over the value of 0.05 and dilating approximately 30 pixels outwards (away from the cell) using MATLAB function ‘imdilate’ (B7). The binary mask with the value 0 in the background and the value 1 inside the dilated region was then multiplied to the image in B6 to obtain a cleaned cell image B8 with the intensity of all pixels in the artificial background cleared to zero. Finally, a Gaussian filter using MATLAB function ‘imgaussfilt’ with a 2D Gaussian smoothing kernel with standard deviation value of 3 was applied to the image to obtain a smoothed cell topography (B9). This step does not change the integrated cell volume.

Article Snippet: Finally, a Gaussian filter using MATLAB function ‘imgaussfilt’ with a 2D Gaussian smoothing kernel with standard deviation value 3 was applied to the image to obtain a smoother cell topography ( B9).

Techniques: Fluorescence, Standard Deviation