Despite my efforts, I still see discordant color differences when I place two or more Landsat8 images side-by-side in a mosaic [1]. For example, one image may be lighter than the other or less saturated, please look to the mosaic.
You did not specify the efforts you already made, nor the bands you displayed and the way you diplayed them, which makes difficult to give a precise answer.
For a start: differences can be due to actual differences on the ground, e.g. adjacent images were recorded on different dates, in different seasons or because of atmospheric differences. Step 1 would be to try and have images from the same year and/or same season and do atmospheric correction. If that is not possible or if differnces still remain, it depends on the purpose what you could do. If the purpose is classification, I would classify the images seperately (with their own training sets) and join the results. If you want a mosaic for visual interpretation or for an overview, you coud use histogram adjustment and stretch each band between the same corresponding values of the same band in the adjacent image. In other words: take one reference image, find the minimum and maximum for each band and stretch the corresponding bands of teh adjacent mages between teh sam values. This should give a visually more pleasing result, provided there are no large differences in land cover (and in shape of histogram) between the adjacent scenes, but it could introduce errors if you want to use this in classification. Soem image procesing softwares have algoritms for mosaicing in which color differences can be smoothed.
You did not specify the efforts you already made, nor the bands you displayed and the way you diplayed them, which makes difficult to give a precise answer.
For a start: differences can be due to actual differences on the ground, e.g. adjacent images were recorded on different dates, in different seasons or because of atmospheric differences. Step 1 would be to try and have images from the same year and/or same season and do atmospheric correction. If that is not possible or if differnces still remain, it depends on the purpose what you could do. If the purpose is classification, I would classify the images seperately (with their own training sets) and join the results. If you want a mosaic for visual interpretation or for an overview, you coud use histogram adjustment and stretch each band between the same corresponding values of the same band in the adjacent image. In other words: take one reference image, find the minimum and maximum for each band and stretch the corresponding bands of teh adjacent mages between teh sam values. This should give a visually more pleasing result, provided there are no large differences in land cover (and in shape of histogram) between the adjacent scenes, but it could introduce errors if you want to use this in classification. Soem image procesing softwares have algoritms for mosaicing in which color differences can be smoothed.
through different band combinations you can improve the image interpretation.
- Badges
- Science topic
- Similar topics
- Geoscience
- Atmospheric Sciences
- Climate Science