An image processing device reads a video consisting of a plurality of image frames from a storage device. A stable region along a Y-axis direction in each image frame is determined according to pixel information of an edge row of each image frame. The device then aligns all the image frames according to the stable region along a Y-axis direction in each image frame, and trims all the image frames by cutting additional image regions excepting the stable region along the Y-axis direction in each image frame, to reduce a shaking degree of each image frame along the Y-axis direction. Furthermore, the device reduces a shaking degree of each image frame along an X-axis direction using similar method by reducing the shaking degree along the Y-axis direction. At last, the device displays stable playback of the video consisting of the aligned and trimmed image frames.