An image processing apparatus capable of obtaining a sufficient HDR effect while keeping smoothness of an appearance of a video image in an HDR composition process. A first computation unit finds a first composition ratio of frames of images that are continuously taken at mutually different exposures based on luminance of at least one of the frames of images. A second computation unit finds a second composition ratio of the frames of images based on the luminance difference between the frames of images. A third computation unit finds a third composition ratio as a final composition ratio for compositing the frames of images based on the first composition ratio, the second composition ratio, and shutter speed at a time of taking the frames of images. A composition unit generates one frame of an image by compositing the frames of images based on the third composition ratio.