In my previous message I should have included matrices (it is in fact in the name MATlab). Anyway, as an image can be considered a m-by-n-by-k matrix, matrix calculus is easily done. You can modify the parameters of the image (contrast, brightness, saturation, histogram equalisation, etc), apply different filters (linear and nonlinear), edge detection, image segmentation, pattern recognition, fast fourier transform, particle counting... If something catches your interest I can explain further (it is quite vast)
Also, matlab supports GPU computing with CUDA-enabled NVIDIA GPUs which is particularly useful for image processing of heavy data (I work with RAW images of 18Megapixel CMOS sensors)