Image processing apparatus, image processing method and non-transitory computer readable medium

An image processing method for change detection using machine learning, the image processing method comprising:

• generating a difference image from input Synthetic Aperture Radar (SAR) images;
• extracting an intensity change feature from the difference image;
• extracting a speckle change feature from input SAR images;
• performing a weighted summation, concatenation or dot product of the intensity change feature and the speckle change feature for each pixel of the difference image to combine the intensity change feature and the speckle change feature to generate a combined feature;
• and
• applying a machine learning algorithm to the combined feature for each pixel to assign the pixel a probability of belonging to each change class and mapping the combined feature to change classes to output a change map.

A non-transitory computer readable medium storing a program for causing a computer to execute an image processing method for change detection using machine learning, the method comprising:

• generating a difference image from input Synthetic Aperture Radar (SAR) images;
• extracting an intensity change feature from the difference image;
• extracting a speckle change feature from the input SAR images;
• performing a weighted summation, concatenation or dot product of the intensity change feature and the speckle change feature for each pixel of the difference image to combine the intensity change feature and the speckle change feature to generate a combined feature;
• and
• applying a machine learning algorithm to the combined feature for each pixel to assign the pixel a probability of belonging to each change class and mapping the combined feature to change classes to output a change map.