Which technique to use for Satellite Imagery?

Hello,

I am new to image classification and am seeking to classify oil spills in SAR imagery using an unsupervised or adaptive thresholding method. I have experimented with various thresholding techniques, including Otsu, but none have consistently performed well across all SAR satellite images. While Otsu's method yields satisfactory results for some images, those containing land areas often obscure the oil spill, as depicted in the attached image.

Additionally, I have attempted mean shift segmentation, but it lacks adaptability, requiring a separate bandwidth setting for each image.

Could anyone provide guidance on which method or combination of methods to employ in order to achieve the desired results?

Thank you

[Ahmad]

anyone can answer?

[Ahmad]

@Image Analyst

Answer 1

[Image Analyst]

After you display the image, call impixelinfo

imshow(sarImage, []);
impixelinfo

then mouse around over the displayed image and see what the values are in the known ground truth region. Then set a threshold to get those areas. Like let's say the values range from 10 to 20, then you'd do

mask = grayImage >= 10 & grayImage <= 20;