Creating masks for objects in an image

I've seen some people who managed to create a mask for objects in an image and I would like to know how it's done. For example, below:

Where four horses are selected and masked as such, where the shapes are very well preserved. Could anyone explain how it's done or is there any matlab functions that allow me to do so?

[Matt J]

Are you looking for an automatic segmentation method, or would a manual boundary tracing tool be enough?

[Stewart Tan]

@Matt J I guess the manual boundary tracing tool? Also, i forgot to mention that i have the coordinates of the blocks where the regions are similar which are the horses. Sorry bout the confusion.

Answer 1

[Image Analyst]

Try the Color Thresholding App in HSV color space. Get everything with blue as one mask. Then make another mask with stuff with low saturation to get the building. Clean up each mask with imfill(), bwareafilt(), or bwareaopen() if needed. Then OR the two masks together and invert. You should be left with the horses.

[Stewart Tan]

@Image Analyst thanks but this image with horses isn't the only one I'm working with. What i do have in possession are the coordinates of blocks of the image that are similar. Is the manual boundary tracing tool that is suggested above used in this case?

[Image Analyst]

You might try this: https://www.mathworks.com/matlabcentral/fileexchange/37197-dem-diffused-expectation-maximisation-for-image-segmentation

Or try segnet if you have the Deep Learning Toolbox.