I want to extract the rendered RGB value and search for similar ones. I want the search option to include RGB color values to find the pixels from the image. For example, a satellite image has pixel value in RGB (154, 246, 165). I want to find the similar pixels from the image. Is it possible?
Your question is not well specified but the answer is yes, especially if you can do some coding yourself. Satellite and color images have particular structures and formats you can find on the web allowing them to be directly accessed pixel by pixel to get the (r, g, b) or color values. You should be able to find utilities or code on the web to input the individual pixel (r, g, b) values. You then have a collection of RGB values and their locations in the image, (x, y, r, g, b) vectors. If you are looking for 'similar' (r, g, b) values (not exact matches) you can use a statistical clustering routine to find them (cluster on the (r , g, b) vectors). How similar is up to you to decide. Alternatively if you want exact matches you can do a brute force search, limiting the search to the particular ranges of the r, g, b values in the image to shorten processing time.
Trivial to do in practically any GIS like QGIS or ESRI ArcMap using a Raster Calculator. Programmatically in Python and R or even on the command line using GDAL, although I recommend using a GIS so you can stack and examine the results immediately.
You should mention your goal, though, because feature extraction is rarely so simple. Shadows for example.
L. J. Hamilton Thanks for mentioning the excellent option. I've tried to find suitable source code to find the rgb values from satellite image, but no result. may be my searching capacity is not that good. could you please mind, if you have any idea, where can I find things such as source code? Obviously I can do clustering analysis in R. But I need to understand the full process of extracting rgb value. Additionally, can I extract the reflectance value of for example cyan (built-up) from the satellite image? I have tried band arithmetic calculation but no tuned result.
If you are using R, it has all the geospatial functions ( https://rspatial.org/ ). See https://datacarpentry.org/r-raster-vector-geospatial/05-raster-multi-band-in-r/ , or google "R language geospatial raster band calculations". I would advise doing some of the tutorials because R requires some explicit housekeeping that a GIS application does behind the scene ( sometimes good, often times bad ).
(1) To understand the process of extracting rgb values from an image you need to look up the specific format for rgb images. RGB files have a header which tells you the height and width of the image, whether or not it is 16-bit, 32-bit etc etc. Even if you find code, app, or executable to extract pixel data automatically it will help to know this. I made a quick search for some code but nil result. I will keep looking because code was readily available when I used to do this kind of work. Michael Patrick's reference to R functions may be a better option. (2) Yes, you can find pure cyan pixels (their RGB values are [0 red, 100 green, 100 blue]) but is that what you want to do? What about pixels that are almost pure cyan? Then you look for pixels within a range of values.
See some good simple C/C++ code at https://smssoftdev-solutions.blogspot.com/2014/07/code-for-reading-bmp-image-files.html Should show all you need if you want to code yourself if you look up image file formats
L. J. Hamilton INRE: 'specific format for rgb images'.
This is basically why the GDAL ( Raster https://gdal.org/ ) and OGR ( vector ) libraries exist, there are a bewildering number of formats and file structures: https://gdal.org/drivers/raster/index.html and https://gdal.org/drivers/vector/index.html
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing