This paper presents a two-stagemethod for detection and segmentation of tumors indigital mammograms employing wavelet-based sub-band images decomposition.In this method,only thehigh frequency subbands information is considered fortumors,which is different from most existing wavelet-based methods that are based on the decompositionlow-low frequency subband.The mammograms arefirst decomposed into different frequency subbandsby down-sampling and suppressing the low-frequencysubband.The image high-frequency features are pre-served by a detection system that employs a suitableimage transform,which can localize the signal char-acteristics in the original and the transform domain.The tumors correspond to high-frequency componentsof the image spectrum,and then we compute the in-verse wavelet transform.Then,a segmentation crite-rion is adopted - a threshold method is used,whichrelies on a combination of criteria used by median fil-tering and segmentation tumors in mammograms fromthe subband images containing only high-frequencies.By appropriate selection of the wavelet basis the de-tection of tumors in the relevant size range demon-strates the efficiency of the method.