Anúncio: todos os projetos não comerciais registrados para usar o Earth Engine antes de 15 de abril de 2025 precisam verificar a qualificação não comercial para manter o acesso ao Earth Engine.
As operações matemáticas de imagem podem ser realizadas usando operadores como add() e
subtract(), mas para cálculos complexos com mais de dois termos, a
função expression() é uma boa alternativa. Consulte as seções a seguir para mais informações sobre operadores e expressões.
Operadores
Os operadores matemáticos executam operações aritméticas básicas nas bandas de imagem. Elas recebem duas entradas:
duas imagens ou uma imagem e um termo constante, que
é interpretado como uma imagem constante de banda única sem pixels mascarados. As operações são realizadas
por pixel para cada banda.
Como exemplo básico, considere a tarefa de calcular o índice de vegetação por diferença normalizada (NDVI, na sigla em inglês) usando imagens do VIIRS, em que os operadores add(), subtract() e divide() são usados:
Apenas a interseção de pixels sem máscara entre as duas entradas é
considerada e retornada como sem máscara. Todas as outras são mascaradas. Em geral, se uma das entradas tiver apenas
uma banda, ela será usada em todas as bandas da outra entrada. Se as entradas tiverem o mesmo
número de bandas, mas não os mesmos nomes, elas serão usadas em pares na ordem natural. As
bandas de saída são nomeadas com base na entrada mais longa das duas ou, se tiverem o mesmo comprimento, na
ordem da primeira entrada. O tipo dos pixels de saída é a união dos tipos de entrada.
O exemplo a seguir de subtração de imagem com várias bandas demonstra como as bandas são combinadas
automaticamente, resultando em um "vetor de mudança" para cada pixel de cada banda simultânea.
Na segunda parte deste exemplo, a diferença ao quadrado é calculada usando
image.pow(2). Para conferir a lista completa de operadores matemáticos que processam
aritmética básica, trigonometria, exponenciação, arredondamento, conversão, operações bitwise
e muito mais, consulte a documentação da API.
Expressões
Para implementar expressões matemáticas mais complexas, use
image.expression(), que analisa uma representação de texto de uma operação matemática.
O exemplo a seguir usa expression() para calcular o índice de vegetação aprimorado (EVI):
O primeiro argumento para expression() é a representação textual da operação matemática. O segundo argumento é um dicionário em que as chaves são nomes de variáveis usadas na expressão, e os valores são as bandas de imagem para as quais as variáveis precisam ser mapeadas. As bandas na imagem podem ser chamadas de b("band name") ou
b(index), por exemplo, b(0), em vez
de fornecer o dicionário. As bandas podem ser definidas a partir de imagens diferentes da entrada ao usar o dicionário do mapa de bandas. expression() usa a "divisão inteira", que
descarta o restante e retorna um número inteiro quando dois números inteiros são divididos. Por exemplo,
10 / 20 = 0. Para mudar esse comportamento, multiplique um dos operandos por
1.0: 10 * 1.0 / 20 = 0.5. Apenas a interseção de pixels sem máscara
é considerada e retornada como sem máscara quando as bandas de mais de uma imagem de origem são
avaliadas. Os operadores de expressão compatíveis estão listados na tabela a seguir.
[null,null,["Última atualização 2025-07-25 UTC."],[[["\u003cp\u003eEarth Engine provides tools for performing image math, including operators for basic arithmetic and the \u003ccode\u003eexpression()\u003c/code\u003e function for complex computations.\u003c/p\u003e\n"],["\u003cp\u003eOperators like \u003ccode\u003eadd()\u003c/code\u003e, \u003ccode\u003esubtract()\u003c/code\u003e, and \u003ccode\u003edivide()\u003c/code\u003e enable pixel-wise calculations between images or an image and a constant.\u003c/p\u003e\n"],["\u003cp\u003eThe \u003ccode\u003eexpression()\u003c/code\u003e function allows implementing custom formulas by parsing text representations of mathematical operations and mapping variables to image bands.\u003c/p\u003e\n"],["\u003cp\u003eWhen using \u003ccode\u003eexpression()\u003c/code\u003e, ensure to handle integer division appropriately by multiplying one operand by \u003ccode\u003e1.0\u003c/code\u003e to preserve decimal values if needed.\u003c/p\u003e\n"],["\u003cp\u003eBoth operators and expressions automatically handle band matching and masking, considering only unmasked pixels in the calculations.\u003c/p\u003e\n"]]],[],null,["# Mathematical Operations\n\n|---------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------|\n| [Run in Google Colab](https://colab.research.google.com/github/google/earthengine-community/blob/master/guides/linked/generated/image_math.ipynb) | [View source on GitHub](https://github.com/google/earthengine-community/blob/master/guides/linked/generated/image_math.ipynb) |\n\nImage math can be performed using operators like `add()` and\n`subtract()`, but for complex computations with more than a couple of terms, the\n`expression()` function provides a good alternative. See the following sections\nfor more information on [operators](#operators) and\n[expressions](#expressions).\n\nOperators\n---------\n\nMath operators perform basic arithmetic operations on image bands. They take two inputs:\neither two images or one image and a constant term, which\nis interpreted as a single-band constant image with no masked pixels. Operations are performed\nper pixel for each band.\n\nAs a basic example, consider the task of calculating the Normalized Difference Vegetation\nIndex (NDVI) using VIIRS imagery, where `add()`, `subtract()`,\nand `divide()` operators are used:\n\n### Code Editor (JavaScript)\n\n```javascript\n// Load a VIIRS 8-day surface reflectance composite for May 2024.\nvar viirs202405 = ee.ImageCollection('NASA/VIIRS/002/VNP09H1').filter(\n ee.Filter.date('2024-05-01', '2024-05-16')).first();\n\n// Compute NDVI.\nvar ndvi202405 = viirs202405.select('SurfReflect_I2')\n .subtract(viirs202405.select('SurfReflect_I1'))\n .divide(viirs202405.select('SurfReflect_I2')\n .add(viirs202405.select('SurfReflect_I1')));\n```\nPython setup\n\nSee the [Python Environment](/earth-engine/guides/python_install) page for information on the Python API and using\n`geemap` for interactive development. \n\n```python\nimport ee\nimport geemap.core as geemap\n```\n\n### Colab (Python)\n\n```python\n# Load a VIIRS 8-day surface reflectance composite for May 2024.\nviirs202405 = (\n ee.ImageCollection('NASA/VIIRS/002/VNP09H1')\n .filter(ee.Filter.date('2024-05-01', '2024-05-16'))\n .first()\n)\n\n# Compute NDVI.\nndvi202405 = (\n viirs202405.select('SurfReflect_I2')\n .subtract(viirs202405.select('SurfReflect_I1'))\n .divide(\n viirs202405.select('SurfReflect_I2').add(\n viirs202405.select('SurfReflect_I1')\n )\n )\n)\n```\n| **Note:** the normalized difference operation is available as a shortcut method: [`normalizedDifference()`](/earth-engine/apidocs/ee-image-normalizeddifference).\n\nOnly the intersection of unmasked pixels between the two inputs are\nconsidered and returned as unmasked, all else are masked. In general, if either input has only\none band, then it is used against all the bands in the other input. If the inputs have the same\nnumber of bands, but not the same names, they're used pairwise in the natural order. The\noutput bands are named for the longer of the two inputs, or if they're equal in length, in the\nfirst input's order. The type of the output pixels is the union of the input types.\n\nThe following example of multi-band image subtraction demonstrates how bands are matched\nautomatically, resulting in a \"change vector\" for each pixel for each co-occurring band.\n\n### Code Editor (JavaScript)\n\n```javascript\n// Load a VIIRS 8-day surface reflectance composite for September 2024.\nvar viirs202409 = ee.ImageCollection('NASA/VIIRS/002/VNP09H1').filter(\n ee.Filter.date('2024-09-01', '2024-09-16')).first();\n\n// Compute multi-band difference between the September composite and the\n// previously loaded May composite.\nvar diff = viirs202409.subtract(ndvi202405);\nMap.addLayer(diff, {\n bands: ['SurfReflect_I1', 'SurfReflect_I2', 'SurfReflect_I3'],\n min: -1,\n max: 1\n}, 'difference');\n\n// Compute the squared difference in each band.\nvar squaredDifference = diff.pow(2);\nMap.addLayer(squaredDifference, {\n bands: ['SurfReflect_I1', 'SurfReflect_I2', 'SurfReflect_I3'],\n min: 0,\n max: 0.7\n}, 'squared diff.');\n```\nPython setup\n\nSee the [Python Environment](/earth-engine/guides/python_install) page for information on the Python API and using\n`geemap` for interactive development. \n\n```python\nimport ee\nimport geemap.core as geemap\n```\n\n### Colab (Python)\n\n```python\n# Load a VIIRS 8-day surface reflectance composite for September 2024.\nviirs202409 = (\n ee.ImageCollection('NASA/VIIRS/002/VNP09H1')\n .filter(ee.Filter.date('2024-09-01', '2024-09-16'))\n .first()\n)\n\n# Compute multi-band difference between the September composite and the\n# previously loaded May composite.\ndiff = viirs202409.subtract(ndvi202405)\n\nm = geemap.Map()\nm.add_layer(\n diff,\n {\n 'bands': ['SurfReflect_I1', 'SurfReflect_I2', 'SurfReflect_I3'],\n 'min': -1,\n 'max': 1,\n },\n 'difference',\n)\n\n# Compute the squared difference in each band.\nsquared_difference = diff.pow(2)\n\nm.add_layer(\n squared_difference,\n {\n 'bands': ['SurfReflect_I1', 'SurfReflect_I2', 'SurfReflect_I3'],\n 'min': 0,\n 'max': 0.7,\n },\n 'squared diff.',\n)\ndisplay(m)\n```\n\nIn the second part of this example, the squared difference is computed using\n`image.pow(2)`. For the complete list of mathematical operators handling\nbasic arithmetic, trigonometry, exponentiation, rounding, casting, bitwise operations\nand more, see the [API documentation](/earth-engine/apidocs).\n\nExpressions\n-----------\n\nTo implement more complex mathematical expressions, consider using\n`image.expression()`, which parses a text representation of a math operation.\nThe following example uses `expression()` to compute the Enhanced\nVegetation Index (EVI):\n\n### Code Editor (JavaScript)\n\n```javascript\n// Load a Landsat 8 image.\nvar image = ee.Image('LANDSAT/LC08/C02/T1_TOA/LC08_044034_20140318');\n\n// Compute the EVI using an expression.\nvar evi = image.expression(\n '2.5 * ((NIR - RED) / (NIR + 6 * RED - 7.5 * BLUE + 1))', {\n 'NIR': image.select('B5'),\n 'RED': image.select('B4'),\n 'BLUE': image.select('B2')\n});\n\nMap.centerObject(image, 9);\nMap.addLayer(evi, {min: -1, max: 1, palette: ['a6611a', 'f5f5f5', '4dac26']});\n```\nPython setup\n\nSee the [Python Environment](/earth-engine/guides/python_install) page for information on the Python API and using\n`geemap` for interactive development. \n\n```python\nimport ee\nimport geemap.core as geemap\n```\n\n### Colab (Python)\n\n```python\n# Load a Landsat 8 image.\nimage = ee.Image('LANDSAT/LC08/C02/T1_TOA/LC08_044034_20140318')\n\n# Compute the EVI using an expression.\nevi = image.expression(\n '2.5 * ((NIR - RED) / (NIR + 6 * RED - 7.5 * BLUE + 1))',\n {\n 'NIR': image.select('B5'),\n 'RED': image.select('B4'),\n 'BLUE': image.select('B2'),\n },\n)\n\n# Define a map centered on San Francisco Bay.\nmap_evi = geemap.Map(center=[37.4675, -122.1363], zoom=9)\n\n# Add the image layer to the map and display it.\nmap_evi.add_layer(\n evi, {'min': -1, 'max': 1, 'palette': ['a6611a', 'f5f5f5', '4dac26']}, 'evi'\n)\ndisplay(map_evi)\n```\n\nObserve that the first argument to `expression()` is the textual representation of\nthe math operation, the second argument is a dictionary where the keys are variable names used\nin the expression and the values are the image bands to which the variables should be\nmapped. Bands in the image may be referred to as `b(\"band name\")` or\n`b(index)`, for example `b(0)`, instead\nof providing the dictionary. Bands can be defined from images other than the input when using\nthe band map dictionary. Note that `expression()` uses \"floor division\", which\ndiscards the remainder and returns an integer when two integers are divided. For example\n`10 / 20 = 0`. To change this behavior, multiply one of the operands by\n`1.0`: `10 * 1.0 / 20 = 0.5`. Only the intersection of unmasked pixels\nare considered and returned as unmasked when bands from more than one source image are\nevaluated. Supported expression operators are listed in the following table.\n\n| Type | Symbol | Name |\n|----------------|---------------------|----------------------------------------------------|\n| **Arithmetic** | + - \\* / % \\*\\* | Add, Subtract, Multiply, Divide, Modulus, Exponent |\n| **Relational** | == != \\\u003c \\\u003e \\\u003c= \\\u003e= | Equal, Not Equal, Less Than, Greater than, etc. |\n| **Logical** | \\&\\& \\|\\| ! \\^ | And, Or, Not, Xor |\n| **Ternary** | ? : | If then else |\n[Operators for `expression()`]"]]