Choosing Transfer Function Settings

To convert an image from pixel values to a displayable map of brightness, Mira uses a transfer function and a palette. The transfer function is described here. A transfer function uses a mathematical prescription that involves three parameters: "Stretch", "Range", and "Sample". The settings for each of these parameters may be changed in Mira.

Choosing Settings for Different Types of Objects

The transfer function settings that works best depends on type of object you are viewing as well as some details about the image itself. What may work best for one class of object may not work so well for another but, thankfully, the difference depends mostly on the type of object in the image. Therefore you can use similar parameter settings based on the object and not have to fret over image statistics and other details. For example, the transfer function that works best for an image of a star field is usually not the best for viewing an image of a galaxy nebula. And neither of these methods usually works very well for viewing an image of a planet or the moon. Therefore, Mira provides many settings for adjusting the transfer function as needed. When you get settings just right for a type of object, it is usually good to save them as a "profile" (see below) so that, in the future, you can re-use the same settings for a similar class of image. The profile keeps all parameter settings organized under a name. For example, you may create a certain group of settings that works well for viewing star fields and save that profile with the name "Start Field". Similarly, you may get an adjustment that works well for the moon (or for the surface of Mars or Ganymede) and save it as "Moon". One that works well for Planets might be saved as "Planet". You can have any number of such profiles and you can refine and re-save them at any time using the Profile Control on the Transfer Function Properties dialog. One of these gets selected as the default transfer function for opening new images. You should select one that you want to use for typical images and make it the default by clicking the [>> Default] button on the Transfer Function Properties dialog.

Mira is supplied with a number of transfer function profiles suitable for a wide range of images. You can use these as is, or use them as a starting point and modify them for your own needs. Here are some guidelines for choosing settings for displaying images of different types of subjects. The settings themselves are described in detail further below.

Understanding Transfer Function Settings

The collection of settings that define a transfer function are described above in general terms. Below, we cover the meaning of these adjustments in a lot more detail. The transfer function is defined by three main parameters: Sample, Range, and Stretch. Each of these is described separately below. The values of these parameters may be adjusted using the Transfer Function Properties dialog.

Sample

This property determines which portion of the image is used to calculate the transfer function. Since the transfer function is computed from the image values, this property controls which pixel values go into the calculation. Initially you may want use the Entire Image to set the scaling so that all parts of the image have a chance to vote on how visible they will be. But later you may discover that that despite what other properties you set, letting all pixels vote does not give the desired result for viewing a specific region of interest. In that case, use the Image Cursor to outline the region of interest and switch to the Cursor Region setting.

Range

This property describes how the minimum and maximum image values will be chosen to set the limit on the transfer function. Only pixels within this range will be assigned bins, or palette indices for display purposes. This means that pixels outside this range will be uniformly "black" at the bottom of the scale or uniformly "white" at the top of the scale, with no visible detail.

The range setting is a sort of contrast value based, for many of the options, upon the histogram (distribution) of pixel luminance. The closer together are the minimum and maximum values, the more rapid the change from "black" to "white" in the screen display and hence the higher the contrast appears to be.

Mira provides a number of percentile options that use the image histogram of pixels inside the Sample region to compute the range values. In the descriptions below, the terms "black" and "white" are used to describe values that are of uniform representation above or below the limits of the transfer function. If a pseudocolor palette is applied to the image, then these values may be pink and periwinkle, but the result is the same that all pixels above or below the transfer limits are of uniform color and lack all detail.

• The Autoscale setting lets Mira choose the minimum and maximum levels based upon its interpretation of the image histogram. When using this method, you must choose a Contrast setting to apply to Mira's choice. The lower the contrast setting, the lower will be the image contrast and vice versa.

• The Min/Max setting accommodates all pixels of any brightness that are found in the sampling region. This is intended to show the whole range of brightness but, for noisy images with a few excessively bright or dark pixels, the result of this setting can be very low contrast. For noisy images, you might be able to overcome this limitation of the method by choosing a sample region that excludes the deviant pixels.

• The 95%, 99% and other percent settings use pixel values only within the 95-th or other percentile of the histogram. These settings therefore allow the brightest pixels to "burn out" in the displayed image. Using a higher percentile setting approaches the same result as the Min/Max setting.

• The Percentile setting lets you specify the histogram percentiles when one of the other built-in settings does not suffice.

• The Specify Z setting allows you to specify the Z value, or image luminance at the extremes of the transfer function histogram. All pixel values below the minimum appear "black" and all those above the maximum appear "white".

Stretch

The Stretch parameter describes how pixel values inside the Range are assigned to different bins. By crowding together or spreading out the values assigned to each bin, you can change the contrast within different portions of the brightness range.

• The Linear option sets equal steps of image luminance for all bins. This means, for example, that the feature contrast at low brightness levels will be the same as the feature contrast at high brightness levels.

• The Logarithmic option compresses the luminance values at the dark end and spreads out the intervals at the brighter end of the image histogram. This gives a faster rate of change of bin index per luminance value at the darker end so the contrast of dark features is enhanced. The rate of change at the bright end is very low, so the feature contrast is reduced at the bright end of the histogram. This setting is a good choice for images where you want to see a lot of contrast in faint background features at the expense of details in bright features.

• The Gamma option is a very versatile one that lets you shift the contrast bias to higher or lower levels or anywhere between. A Gamma value less than 1 shifts the higher contrast to lower luminance and a gamma value greater than 1 shifts the higher contrast to higher luminance. A gamma value of exactly 1.0 gives the same result as a Linear stretch.

Related Topics

Transfer Functions, Transfer Function Properties Dialog