This invention maintains pixels as a source of a high frequency component such as a character and generates the encoded data having a fixed length for one block. A block generator reads out image data of one block from image data. An extractor extracts a color of a pixel having high frequency component, and generates information identifying pixels of the extracted color and those of non-extracted color. A first encoder lossless-encodes the identification information. A detector detects the data size of the generated encoded data. A substitution unit calculates the average value of colors of the pixels having the non-extracted color, substitutes the color of the pixels having the extracted color by the average value. A second encoder encodes the pixels after the substitution. An adjustment unit adjusts the data size of the encoded data from the second encoder in accordance with the code size detected by the detector.