Would a local operator perform luminance compression better than a global one? One thing to take note of when working with a local operator is that, when looking at a small window of values in an image, if that window happens to encompass an edge between two surfaces with greatly varying intensities, and one weights the values only with respect to their positions, then values across this edge may have a negative effect on the mapped pixel. For example, if one wishes to map a bright pixel close to such an edge with dark pixels on the other side, and a window used takes into account the values across that edge, then the influence of the dark pixels on the averaging of the local area would be disproportionate, causing the bright pixel to appear much darker than one might expect it to be. The same thing would happen to the dark pixels across the edge when they are mapped, causing them to be mapped much brighter than one might expect. This effect is usually manifested as halos across sharp boundaries of intensity. However, if values with extreme deviations can be ignored from the averaging, then the final mapping would more accurately replicate the original scene in reducing the advent of halos, caused by averages influenced by extreme values. This is the motivation for a gain controlled mapping method that attempts to weight only those neighboring pixels with similar intensities to the pixel we wish to map, as suggested by Pattanaik [1].

The idea is that surrounding values which are too much brighter or darker than the pixel being mapped should be ignored from the weighing function in mapping a pixel. For example, if one defines “too bright” or “too dark” to be a factor of 5 from the current pixel being analyzed, then neighbors whose values are not within the range can be ignored:

Figure 1

Where Ic is the intensity of every pixel and I the intensity of neighboring pixels.

Since such edges in nature have a smoother profile than just an edge, a smoother profile was suggested with the formula:

Figure 2

We made the contrast threshold (the “how bright is too bright?” parameter) adjustable to test the effects of different values. Below is an example used to test the effect of this parameter. For cases where the contrast threshold is less than the contrast ratio between two areas across an edge, the filtered result is a sharp boundary between the two surfaces. For cases where the threshold parameter is greater than the contrast ratio between the surfaces across a sharp boundary, the filtered image produces blurred versions of the boundary. For the image with threshold level 2, there are obvious “clipping” effects around areas with fairly sharp boundaries, such as the ceiling structures. These effects quickly disappear with increasing threshold values up to 5. Beyond that, there is little difference in the final image with respect to the threshold value, which seems to suggest that white is about 5 times the intensity of mid gray while black is 1/5 the intensity of mid gray. Another tunable parameter is the radius used in the local averaging. The actual size of the local window is a square with sides of length 2*radius+1. Increasing the radius produced blurrier and more homogenous images, while decreasing it sharpened the image while reducing the effect of the averager. Why is this case? More values are averaged with larger radii and thus a more homogenous mapping is assigned, which makes the mapped pixel closer in intensity to its neighbors.

For mostly continuous surfaces with little change in intensity throughout its range, the adaptive gain control operator averages out these low contrast areas, which explains the blur in the output images However, this type of blur is undesirable for most photographic applications, since a significant amount of detail is sacrificed.

Figure 3: Threshold = 2, Radius = 2

Figure 4: Threshold = 5, Radius = 2

Figure 5: Threshold = 10000, Radius = 2

Figure 6: Threshold = 5, Radius = 10

Figure 7: Threshold = 5, Radius = 1

Content actions

Add module to:

My Favorites Login Required (?)

'My Favorites' is a special kind of lens which you can use to bookmark modules and collections. 'My Favorites' can only be seen by you, and collections saved in 'My Favorites' can remember the last module you were on. You need an account to use 'My Favorites'.

| A lens I own Login Required (?)

Definition of a lens

Lenses

A lens is a custom view of the content in the repository. You can think of it as a fancy kind of list that will let you see content through the eyes of organizations and people you trust.

What is in a lens?

Lens makers point to materials (modules and collections), creating a guide that includes their own comments and descriptive tags about the content.

Who can create a lens?

Any individual member, a community, or a respected organization.

What are tags?

Tags are descriptors added by lens makers to help label content, attaching a vocabulary that is meaningful in the context of the lens.

| External bookmarks

Footer

This work is licensed by Tianhe Yang, Taylor Johnson, Sarah McGee, and Robert Ortman under a Creative Commons Attribution License (CC-BY 2.0), and is an Open Educational Resource.

Last edited by Taylor Johnson on Dec 25, 2006 4:22 pm -0600.