According to the method and apparatus of processing a digital image for noise suppression and sharpness enhancement, sharpness enhancement and smoothing are performed on an original image to obtain a sharpness enhanced image and a smoothed image, from which image data containing subject image edges and noise (graininess) is determined; edge detection is performed from the original image to determine weighting data for a noise (or grainy) region, which is multiplied by the above image data to determined noise (grain) data of each color in a noise region, from which a black-and-white noise (grain) component and a color noise (dye grain) component are discriminated and separated based on a color correlation component obtained by calculating a color correlation or an obtained local grain coefficient representing a spatial magnitude and a variation magnitude of density variations due to graininess; the thus obtained black-and-white noise component and color noise component are multiplied by their suppressing coefficients to determine a black-and-white noise suppressing component and a color noise suppressing component, which are selectively removed from the sharpness enhanced image data, whereupon a processed image in which the noise suppression is achieved while retaining the sharpness enhancement in the edge region of the image is created. Consequently, graininess can be suppressed while enhancing image sharpness, without causing the unnatural artifacts which may be produced by the discoptinuousness of the boundary between a region from which graininess has been removed and a region where sharpness enhancement has been done.