An image dynamic range compression method for an image processing device is provided. The method includes receiving a plurality of input pixels which include a first input pixel, obtaining a luminance pixel value of each of the input pixels, executing a filter execution according to the luminance pixel values of the input pixels to obtain a filter result corresponding to the first input pixel, transforming the filter result into a function result according to a function, and obtaining an output luminance pixel value of the first input pixel according to the luminance pixel value of the first input pixel and a transforming ratio between the function result and the filter result. Accordingly, the method reserves the details in an image and color information, and increases the execution speed.