.png)
Advanced Sharpen
The Advanced Sharpen transformation is a powerful tool for sharpening digital images.
The Advanced Sharpen dialog box consists of a fixed upper panel and a variable bottom panel. At the bottom of the dialog box are three tabs that let you select the three phases – Noise Reduction, Speck Removal or Sharpening. Clicking the corresponding tab causes the lower part of the dialog box to show the controls for the phase you selected. The 1:1 preview image always displays the results of the current phase and its preceding phases only, so you are only seeing the final preview when you select the Sharpening tab. These three operations are applied sequentially to the input image to produce the output image. You can adjust the size of the 1:1 preview image by resizing the dialog box. Clicking on the input image display centers the preview on the location you clicked on, or you can click and drag to scroll the preview window.
Noise Reduction
In this phase, the image is blurred to reduce noise without softening edges. This feature is useful for reducing film grain artifacts in scanned images or noise in digital camera images, especially in smooth regions such as clouds or clear skies. The blurring algorithms incorporate thresholding techniques to prevent softening sharp edges and other high contrast detail.
Speck Removal
Isolated pixels or small clusters of pixels significantly lighter or darker than the surrounding area are called specks. Specks can be caused by dust or imperfections in the film, camera or scanner. While the noise reduction phase does not remove specks, it does make them easier to detect by smoothing the area around them. Several parameters can be adjusted to control the speck removal process so you can eliminate specks while preserving fine image detail.
Sharpening
Once the preceding two phases are complete, the image is ready to be sharpened. Since sharpening amplifies noise and exaggerates specks, the preceding two phases make it possible to sharpen images that might otherwise break up when sharpened. The sharpening phase can be controlled using selectable thresholds to sharpen high contrast detail without amplifying low contrast detail.
General
Clicking on the input image re-centers the preview display on the part of the image you clicked on.
Shift-Clicking on the input image window displays a red, vertical marker in the histogram display (see below) for the current phase until you release the mouse button. The position of the marker corresponds to the histogram value for the part of the image at the current cursor location. If you drag the cursor over the input image with the shift button down, the range of red marker positions is recorded in the white bar just below the histogram. To clear this, click on the input image with the shift button raised.
To bypass a phase, just click one of the other tabs leaving the settings for the phase you want to skip defaulted to take no action. Thus, you can use any of the three parts of Advanced Sharpen without the others.
Noise Reduction
.png)
Noise reduction is performed one pixel at a time using modified Gaussian blurring. Gaussian blurring works by replacing each pixel with a weighted sum of the pixels in a roughly circular region surrounding it. Nearby pixels are assigned greater weights than distant pixels following a Gaussian distribution curve.
There are several noise reduction parameters you can control.
Blur Radius
The larger the radius, the broader the distribution, and the more surrounding pixels are included in the neighborhood, resulting in a blurrier image. Using a larger radius also increases the amount of computation and consequently runs slower.
Blur Amount
This setting controls how much of the selected blur is applied to the image. When set to zero, no blurring is performed -- as the amount increases, blurring is also increased. This is done by computing a weighted average between the original image and the fully blurred image.
Histogram
The histogram display just below the Blur Amount slider lets you perform selective blurring by including only those neighboring pixels in the computations that differ by less than a given threshold amount from the pixel you are sharpening.
Unlike histograms used in the Brightness Curve transformation, the Noise Reduction histogram reflects not the tonality of the image, but rather the roughness of its texture. Bars at the left end of the histogram correspond to pixels in very smooth areas of the image, i.e. areas of nearly solid color.
Bars further to the right in the histogram indicate the presence of pixels whose neighbors exhibit more variation. Pixels that are near prominent edges or in areas that have a lot of fine detail will thus contribute to histogram bars to the right of pixels that are in the middle of a smooth textured feature such as a cloud or clear sky. By distinguishing between these two types of pixels, Advanced Sharpen can smooth areas of low detail while leaving edges and image details intact.
Traditional Gaussian blurring does a very good job of smoothing out noise, but also softens edges and smooths out fine detail. To prevent this, Picture Window excludes all pixels from the blurring computation that differ by more than a specified threshold from the central pixel. Consider for example two areas of more or less solid color that meet at a sharp edge.
When blurring pixels on one side of the edge, if the threshold is set properly, only similar neighbors are included in the weighted sum, thus preventing pixels from the other side of the edge from being included in the computation (and thus blurring) the edge.
Rather than implementing a sharp cutoff, Picture Window lets you set two thresholds -- one below which all pixels are included and one above which all pixels are excluded. Between the two thresholds, pixels make a partial contribution to the blurring computations. This method helps reduce artifacts from the sudden switchover.
To let you make an intelligent decision about how to set the thresholds, Picture Window computes and displays a histogram of the differences between nearby pairs of pixels in the image. Markers are displayed below the histogram to indicate the current threshold settings, and you can adjust these by dragging the markers. Also, when you click and drag on the input image, the corresponding location in the histogram is highlighted so you can see what difference values are present in each part of your image.
For best results, try to apply just enough blurring to remove noise while preserving as much image detail as possible -- this usually means setting the radius and thresholds to the minimum values that still eliminate noise in smooth areas of the image.
Speck Removal
.png)
Isolated single pixels or small groups of pixels that are either darker or lighter than the surrounding pixels are called specks. Since they represent a large local variation from their neighbors, specks are not removed by the noise reduction phase.
Speck removal operates by first detecting and then removing specks from the results of the preceding noise reduction phase. There are several settings to help you control the process.
Speck Size
This setting controls how large a speck will be removed. Try to use the smallest value that removes the specks you want to eliminate. Setting this value too large may cause a loss of image detail. Speck detection works by looking at a region twice the diameter of the speck size. It first computes the average value in the ring of pixels surrounding the speck and then compares this to the pixels in the center.
Remove Light/Remove Dark
These settings tell the speck detector to search for specks that are lighter and/or darker than the background. If neither checkbox is checked, no action is taken. If your specks are primarily black (as is the case with dust on scanned positives), select Remove Dark. If your specks are primarily white (as for dust on scanned color negatives), select Remove White. If you have both kinds of specks, select both.
Speck Difference Threshold
This slider tells the speck detector to ignore specks that are lighter or darker than the surrounding area by less than a given threshold amount. Set this slider as high as possible to remove specks without removing too much image detail.
Sharpening
.png)
Once blurring and speck removal are completed, you may find that relatively little sharpening is needed. In any case, there are several settings you can use to control how much and what parts of the image is sharpened. Sharpening is performed by first blurring the image and then comparing the original image to the blurred version and exaggerating the differences between them. Thus pixels that stand out from their neighbors will stand out even more in the result image while pixels that are close to the average of their neighbors will be changed relatively little.
Sharpen Radius
This setting controls how much the image will be blurred in preparation for sharpening. For best results, this setting should be smaller for images that are already relatively sharp and larger for images that are softer.
Sharpen Amount
This setting controls how much the differences between pixels in the original image and the blurred image are exaggerated. Higher settings increase the sharpening effect, but if you oversharpen you will start to see light halos around dark objects and dark halos around light objects. These halos can be reduced however by setting the Max Overshoot/Undershoot (see below).
Max Overshoot/Undershoot
These settings are useful for controlling light or dark halos that occur around distinct edges. The white slider limits the maximum amount a pixel can be lightened when sharpening and the black slider limits how much it can be darkened. Setting max overshoot or undershoot to zero suppresses all sharpening while setting them to 255 performs no halo suppression. Use the highest value that produces acceptable haloing.
Histogram
The histogram display just below the Sharpen Amount slider lets you perform selective sharpening by including only those neighbor pixels in the computations that differ by more or less than a given threshold amount from the pixel you are sharpening. Raising the lower limits helps prevent amplifying fine texture or noise. Lowering the upper limits helps reduce halos.
Settings Menu
.png)
No/Medium/High Histogram Expansion
These settings let you control how much expansion is applied to the histogram display. The default setting of no expansion displays histograms normally. The more expansion you select, the more small histogram values are boosted. Histogram expansion can make it easier to see small variations that would otherwise be swamped by large peaks.
Blur All/Chrominance Only
This setting applies only to the noise reduction phase – it determines how blurring is performed. Normally, the red, green and blue channels are all blurred equally. Optionally, the image can be split into two parts: a brightness component and a pair or color components collectively referred to as chrominance. By blurring only the chrominance, brightness variations remain unchanged while the color information is blurred. This type of blurring can be helpful for reducing color noise in digital camera images.
Sharpen All/Luminance Only
This setting applies only to the sharpening phase – it determines how sharpening is performed. Normally, the red, green and blue channels are all sharpened equally. Optionally, the image can be split into two parts: a brightness or luminance component and a pair of color components. By sharpening only the luminance, color variations remain unchanged while the brightness information is sharpened. This can prevent certain subtle false coloring effects from occurring at sharp edges.
Preview
A 1:1 preview of the sharpen effect is displayed in the right-hand side of the dialog box. Clicking on the input image display centers the preview on the location you clicked on, or you can click and drag to scroll the preview window.
If you enlarge the dialog box, you can make the sliders wider and the preview area narrower or vice versa by dragging the vertical separator line between them.
Preview Button
Clicking the Preview button causes the full output image to be recalculated, not just the preview area. This can be useful if you want to view the entire image at once and not just a small section, but it can take longer to compute. Unlike the preview area, the preview button always shows the results of all three tabs, regardless of which one is currently selected.
Difference Button
Clicking the Difference button causes the preview to display the difference between the original image and the preview image. Differences are exaggerated by a factor of two to make them more apparent.
Tips
Trying to remove too much noise from an image usually results in a loss of detail as well. Usually it is better to under-correct noise as it is often invisible when the image is viewed at a normal scale rather than zoomed in to 1:1.