Character spacing values in a document image are extracted and a variance is calculated for fluctuations in the character spacing values. When the calculated variance is lower than a preset threshold value, the document image is determined as having watermark information embedded therein. Such use of the variance in the character spacing values enables high-speed determination of the presence or absence of character-spacing watermark information. At this time, it is possible to speed up the determination by using only some character spacing values in the document, instead of using all character spacing values.