In an image conversion method and apparatus, luminance Y.sub.3 of each pixel of a converted image is obtained by use of an expression Y.sub.3=C.sub.1Y.sub.1+C.sub.2Y.sub.2 and from luminance Y.sub.1 of a corresponding pixel of a source image, luminance Y.sub.2 of a corresponding pixel of a low frequency image, and C.sub.1 and C.sub.2 which are functions of the luminance Y.sub.2. Since C.sub.1+C.sub.2 is constant when Y.sub.2.ltoreq.T.sub.2, contrast of a low frequency component can be prevented from decreasing. In a portion in which the luminance Y.sub.2 of the low frequency image is low, the low frequency image is enhanced. In a portion in which the luminance Y.sub.2 of the low frequency image is high, the low frequency image is suppressed. In a method and apparatus for detecting noise level of an original signal in real time, a plurality of local regions are set on an input image in such a manner that the local regions are distributed uniformly over the entire area of an image. In each local region, determination as to whether or not luminance is saturated is performed. Variation in luminance is detected in each of a plurality of unsaturated local regions. Noise level is determined on the basis of the detected variation.