Enhancing and color-enhancing satellite imagery


The principle of image enhancement is not related to satellite imagery only, it is a technique of improving digital images in general, dating back to early days of digital image processing. Its main purpose is to show more details in the image, details which might otherwise remain hidden or hard to see (to a human eye). In principle, every digital camera applies some level of image enhancement to convert the raw images as captured by the camera's sensor to a final output image, with most camera users not knowing anything about what their camera does after it takes the picture. Similar techniques can be applied to the satellite imagery, to improve their quality or to increase the amount of information they can show to the end users. For more information on image enhancement in satellite meteorology go the webcast about Meteosat IR Enhancements.

 Color-enhancing IR brightness temperature imagery
When using the IR window thermal band (from now on just "IR") images displayed in black and white non-enhanced form, such as the IR10.8 band images shown in figures 3b, 4b, 5b and 6b, there is not much we can say about the BT field of the storm top, about its lowest values, detailed structure of the storm top, location of the storm core, etc. To overcome this drawback, the IR images can be "color-enhanced" - part of the black and white linear grey scale being replaced by dedicated colors, each color representing a specific BT, or alternatively to a short BT interval. In the first case we speak about a "continuous" or "smooth" color enhancement (palette), in the second case we refer to it as to "block" color enhancement (palette). The principle of the color enhancement technique (using a continuous blue-red color palette) is illustrated below, in Figure 7.
For a full size view, click on the image

Figure 7. Principle of the color enhancement. The IR10.8 image at left is displayed using a grey scale palette, ranging from 320 K (shown in black) to 200 K, represented by white. The temperatures between these margins are assigned to grey shades linearly. At right, we have the same IR10.8 image, however displayed using a color enhancement. Part of the original grey scale from 240 K to 200 K is replaced by a color scale (in this case called blue-red color palette), where the blue color corresponds to 240 K, and dark red color to 200 K. The brightness temperatures between 240 K and 200 K are assigned to specific, unique colors, given in the color scale at bottom-right (with the black markers indicating 10 K intervals). The grey (black-white) part of this enhancement is stretched between 320 K  and 240 K.
As has been already mentioned in the previous section, the typical, average cloud-top temperatures of convective storms depend strongly (among other factors) on the geographical region. For this reason, the example of IR BT color enhancement as shown in Figure 7 will work fine for most of storms over Europe, but will not be as good for storms in the tropics or subtropics. Figure 8 below shows how the appearance of a storm changes with different enhancements. The color palettes used in the examples below always show a range of 40 K, however with different warm and cold end of the enhancement. You can click the images to see them in their full resolution; each of the color-enhanced images contains (at top-right corner) the corresponding color scale:


For a full size view, click on the image
Figure 8a. 2013-05-18, 12:30 UTC, Meteosat-10, Nigeria (the same storm as shown in Figure 6). IR10.8 band, non-enhanced black and white image. All the color-enhanced images below (Fig.8b - Fig.8g) show the same storm and the same IR10.8 band, but each using a slightly modified range for the blue-red color palette; the values of the color-enhancement range are given below each of these images.
Figure 8b. The same storm as in Fig. 8a, using the blue-red color enhancement for BT range 180 K - 220 K. Clouds and land at warmer temperatures (220 K - 320 K) are shown in grey scale.
Figure 8c. The same storm as in Fig. 8a, using the blue-red color enhancement for BT range 185 K - 225 K. Clouds and land at warmer temperatures (225 K - 320 K) are shown in grey scale.
Figure 8d. The same storm as in Fig. 8a, using the blue-red color enhancement for BT range 190 K - 230 K. Clouds and land at warmer temperatures (230 K - 320 K) are shown in grey scale.
Figure 8e. The same storm as in Fig. 8a, using the blue-red color enhancement for BT range 195 K - 235 K. Clouds and land at warmer temperatures (235 K - 320 K) are shown in grey scale.
Figure 8f. The same storm as in Fig. 8a, using the blue-red color enhancement for BT range 200 K - 240 K. Clouds and land at warmer temperatures (240 K - 320 K) are shown in grey scale.



Figure 8g. The same storm as in Fig. 8a, using the blue-red color enhancement for BT range 205 K - 245 K. Clouds and land at warmer temperatures (245 K - 320 K) are shown in grey scale.
As can be seen from the images above, the appearance of the storm top - the amount of details which can be perceived or resolved by a human eye in these - changes significantly with the shift of the color enhancement range. In this particular case, the enhancement range used for the two uppermost images (Fig.8b and Fig.8c) reveals the coldest spots, but the thermal structure of the rest of the storm top remains hidden. On the contrary, the enhancements used for the last two images (Fig.8f and Fig.8g) result in "over-saturation" of large parts of the cold tops by dark red color, making localization of the coldest pixels uneasy or even impossible. The optimum enhancement for this case appears to be the one shown in Fig.8d (190K-230K), which clearly reveals the coldest tops, and also shows relatively well the thermal structure of the warmer parts of the storm top. The warmer part is shown slightly better in the next image, in Fig.8e (195K-235K), but the coldest tops already merge with their nearest surroundings.

 The change of appearance of the storm top with even finer enhancement shifts, at 1K steps, can be followed in the following animation. It shows the same storm as above, looping from 180K-220K scale to 205K-245K scale, and back:



 If we would apply the same exercise to storms at higher latitudes (e.g. for central Europe), we would get a different result - as the most suitable enhancement would turn out to be the one used here in Fig.8f, for the 200K-240K range. For even higher latitudes, such as those of southern Scandinavia, one might have to use the "warmest" enhancement, 205K-245K.
Of course, there are many other ways how to modify the color enhancement - instead of shifting the 40K range as shown above, it can also be stretched or squeezed to a larger/smaller BT range, as needed to show broader or finer BT intervals, depending on the specific use or purposes. However, the order of the color scale (red for cold end, blue for the warm end of color enhancement) should be preserved. Why the red color is being used for the cold temperatures is a topic of the following exercise.
Why color enhanced IR images are used, why they are important, what is the purpose of all of this, will be discussed in the following sections.




Technical note: the blue-red color palette shown above (and used further on, throughout this material) is a palette commonly used in various remote sensing processing/visualization software systems, and was adopted by the Convection Working Group of EUMETSAT as a recommended palette (or a standard one) for color enhancing of the IR BT imagery, namely when displaying convective storms. Further details on this palette, together with sample files for implementing it in various satellite software packages and user stations can be found here.