Goal | Spatial Filtering | Modeling Brightness vs. Depth | Results| Moloka'i Title Page

Underwater Photography showing loss of light and color with water depth. (170 kilobytes)

The main objective of this task was to investigate the removal of water depth effects in digital aerial photographs using both a stand-alone spatial filtering procedure and modeling of brightness values versus the SHOALS Lidar bathymetry data.

TOP

The spatial filtering technique used in this project computes the average brightness value within a user specified window and subtracts it from the brightness value of the pixel at the center of the window, it then assigns this value to the pixel at the center of the window. Our software (USGS MIPS) has the capability to apply a filter size ranging from 3 by 3 to over 2001 by 2001 pixels; the size of the window used in this study was equal to 301 by 301 pixels. Spatial filtering can be used to remove regional low frequencies from digital images, therefore, it can be used to subdue the regional effect of water depth on the brightness seen in clear coastal water photographs. An advantage of the spatial filtering procedure is that it does not require bathymetry data; however, a disadvantage is that only the low frequency or regional trend is removed and not local high frequency bathymetry variations.

TOP

Figure 1: Graph showing the relationship for each of the three colors in the digital aerial mosaic for the area used in the study. (700 x 512, 14 kilobytes)

Figure 2: Scattergrams illustrating the effect that the depth corrections had on the blue and green bands. (600 x 600, 24 kilobytes)

A second correction procedure investigated used the digital airborne SHOALS bathymetry data georeferenced to the digital aerial mosaic. The digital photo mosaic's brightness values of sand at depth ranging from two to sixty feet were extracted for the blue, green, and red colors. The brightness values were used to generate a relationship/model between the blue, green, and red colors and water depth for the given data. The graph (figure 1 on the right) shows the resulting relationship for each of the three colors in the digital aerial mosaic for the area used in the study. It was assumed that sand at different water depths has similar brightness, so changes in digital values were directly related to differences in water depth.

The relationships established between water depth and the digital values in the blue, green, and red colors were used to normalize the aerial photo mosaic brightness on a pixel-by-pixel basis to those in the very shallow waters. The advantage of this procedure is that depth information at each pixel is used, so it corrects for both regional and local high frequency bathymetry variations. However, a disadvantage is that a detailed bathymetric data set is required for the correction.

TOP

Figure 3: Examples of the two corrections performed on a portion of the digitized aerial mosaic. (400 x 578, 128 kilobytes)

The spatial filtering procedure allows a relatively low frequency depth correction to be made without the need of a bathymetry file. However, since it is only a first order relative correction, some color information is lost. Using the pixel-by-pixel bathymetry correction procedure a linear model between water depth and brightness change gave an R-squared correlation of 0.987 for the blue band, while a second order model gave an R-squared value of 0.949 for the green band. The water penetration capability of the red spectral band is much lower than the blue and green bands and its R squared value was 0.795.

TOP

U.S. Department of the Interior | U.S. Geological Survey | Jobs Home | Search | Disclaimers & Privacy Web Rings Send comments to webman@TerraWeb.wr.usgs.gov URL: Last Modified: Thu Oct 31 15:12:10 MST 2002