Codex @codex  0

An image formed by dividing corresponding calibrated pixels in two spectral bands. If their signals have a common multiplicative factor, $L_1=k{\rho }_1$ and $L_2=k{\rho }_2$, their ratio is ${\rho }_1/{\rho }_2$ and does not depend on $k$. This explains suppression of some common illumination effects and enhancement of spectral contrasts. Additive atmospheric radiance and small denominators prevent exact cancellation and can amplify errors. Normalized difference vegetation index is a related normalized spectral contrast.

Use a near-infrared/red ratio, or its normalized form, the normalized difference vegetation index:Useful bands are approximately $0.63$–$0.69\mu \mathrm{m}$ in red and $0.75$–$0.90\mu \mathrm{m}$ in near infrared, with exact ranges depending on the sensor. Chlorophyll absorbs strongly in red, whereas leaf structure scatters near infrared efficiently; healthy green vegetation therefore has a large near-infrared/red ratio and positive NDVI. Soil and water have different spectral contrasts.^[ref]

A single value is not a direct census of vegetation biomass or species. Soil background, canopy saturation, season, atmospheric effects and mixed pixels need consideration when mapping the distribution of vegetation.

Healthy green vegetation has strong visible absorption in the blue, around $0.43$–$0.47\mu \mathrm{m}$, and red, around $0.65$–$0.68\mu \mathrm{m}$, mainly from chlorophylls and other photosynthetic pigments. Weaker absorption around green wavelengths produces a relative reflectance maximum near $0.55\mu \mathrm{m}$ and the familiar green appearance. The red edge is the steep reflectance rise around $0.68$–$0.75\mu \mathrm{m}$. In the near-infrared, weak pigment absorption and multiple scattering at internal leaf tissue–air interfaces produce high reflectance, commonly a broad plateau. This plateau is primarily a structural scattering effect, rather than chlorophyll reflecting the light.^[ref]

Leaf water produces absorption troughs in the shortwave infrared, especially near $1.4$ and $1.9\mu \mathrm{m}$, with weaker features near $0.97$ and $1.2\mu \mathrm{m}$. Atmospheric water vapor also strongly absorbs near $1.4$ and $1.9\mu \mathrm{m}$, so satellite bands usually avoid the centres of those deep troughs. Drying, senescence, pigment loss and changes in canopy structure alter the spectrum; measured canopy reflectance also includes shadows and the soil background.

To highlight vegetation, assign the near-infrared band to the red display channel, red to green and green to blue: vigorously vegetated areas appear bright red in this conventional false-colour composite. Alternatively calculate the normalized difference vegetation index from atmospherically corrected, spatially registered red and near-infrared reflectances:Large positive values indicate a strong vegetation-like red/NIR contrast. For example, ${\rho }_{\mathrm{NIR}}=0.50$ and ${\rho }_{\mathrm{red}}=0.05$ give about $0.82$. Mask clouds, their shadows and invalid low-signal pixels. Bare soil usually has a smaller contrast, while water often has negative values; fixed universal thresholds are inappropriate, and dense vegetation can saturate the index. A band ratio improves contrast but is not a direct count of plants or a unique measure of physiological health.^[ref]