Hot & Cold Pixel Remover
The Huygens software standardly includes a tool to remove (local) Hot Pixels. This tool uses a filter to detect hot pixels and replaces them with the median of their nearest neighbors.
Removal of hot pixels is useful:
- For deconvolution; deconvolving images with hot pixels influence the deconvolution, especially when those hot pixels are not due to strong noise, but to deficiencies of the camera.
- To avoid false colocalization; when hot pixels occur at the same position in each channel, this will affect the colocalization analysis.
- For visualization; hot pixels disturb the contrast range, and could force the contrast of the image to be very low. This may cause that objects of interest with lower intensity become almost invisible.
We have come across artifacts due to the presence of hot pixels a few times. When the image seems black after deconvolution, try to remove hot pixels first and see, after repeating the deconvolution process, if the original contrast of the image is restored.
WindowThe Hot Pixel Remover window can be found in Huygens Essential under Tools and in Huygens Professional under Deconvolution.
At startup the Hot pixel corrector window shows the orthogonal views of the image. At the bottom of the window there are several tabs with which you can change the visualization of the image and of the hot pixels. The options at the Hot Pixel display settings tab are:
- Threshold: Threshold is used to filter hot pixels on a value calculated with an algorithm that takes the intensity and location of the neighboring pixels into account. A high threshold results in a lower count of hot pixels, and a low threshold in a higher count. At start-up the strength is always set to 50.
- Repeating: If the hot pixels originate from an abnormal charge leakage or defect in the CCD, then these defects are most likely visible at the exact same xy-locations for each z-slice, time frame, and very likely also for each channel. If you know that the hot pixels are conserved in a particular dimension, you can set this dimension here. This extra information is then used to detect hot pixels even better. False positives will most likely be eliminated if the information of other z-slices is taken into account. By default the in Z is turned on, because hot pixels are likely to be conserved in every slice of a 3D stack. If you want to remove hot pixels which are a product of deconvolution, all the repeating options should be turned off because in that case you are searching for a single hot pixel which is not conserved in any dimension.
- Results: The results of the hot pixels are displayed in a plot as the hot pixel count per slice for a specific channel. You can switch between the channels. For time series images, the pixel count can be plotted per frame.
|Estimating Threshold & Viewing Result
The strength is set as default to 50. It is difficult to determine a priori what the correct strength settings to separate the real hot pixels from the false positives. To minimize false positives, you can press the Estimate threshold button which will set the strength to an estimated position. To prevent centers of small bright objects to be included as hot pixels, fine tuning with the plus and minus sign may be necessary. If the estimated threshold is higher than 100, the hot pixel remover makes a safe decision and indicates that no hot pixels are detected. The reason that the estimation is not done at startup is because it may be a time-consuming process. Note that every time you change the repeating settings you need to re-run the estimation.
You can zoom in the scene and change contrast and color settings to optimize the visualization. Still, it may be difficult to find hot pixels. Then, it may be useful to right click on the image and use the option Find closest hot pixel, which will move the mouse to the closest detected hot pixel. The Display pull-down menu offers you the possibility to show only the hot pixels, the image or both. This may be useful for setting the threshold.
|Remove hot pixels
Once you are satisfied with the hot pixel detection, you can correct your image. Simply press Correct and a new image will be created with the hot pixels corrected. This image will be shown in the main window of Huygens Professional. The hot pixels are replaced by the median of their neighborhood.
Comparing the resultsYou may use the Twin-slicer to view the difference between the original and the corrected version. Below you find a comparison of the image that is used here as an example. The hot pixels are clearly not present in the corrected image.
Also note that the contrast setting is not the same for both images. For a correct comparison the images are set to show somewhat the same contrast. However, the contrast for the second image can be set lower, because the high values are removed.
When checking a line profile that goes through two hot pixels, the correction is clearly shown. Instead of a high peak the dotted-line shows the correction which straighten the line profile.
Checking the histogramYou may also take a look at the histogram of the image (right click image > Analyze > Histogram or open the Colocalization Analyzer). Because this example images has 2 channels, we can look at the 2D-histogram and we see a clear clean up of unwanted intensities after Hot pixel correction. Click on one of the histograms to switch between before and after. This also shows why there is a gain in contrast after the hot pixels are removed.
Images were used with permission of Dr. Rebecca Lee and Genevieve Phillips, Fluorescence Microscopy Shared Resource, University of New Mexico School of Medicine, Albuquerque, USA