Mathematical morphology has a good application effect in the magnetotelluric(MT) data noise suppression. However, it is still in the blank stage of MT imaging processing. It is based on changing the spatial distribution form or structure of the target in the image to achieve the target extraction. The objective is to achieve the extraction of useful geological information in MT imaging processing. Firstly, the morphological expansion-contraction gradient and morphological closed-open gradient algorithm were used to process the TM and TE mode forward imaging, which effectively improved the reflection of the forward image. The target abnormal body contour was selected. Then, the inversion imaging results of the measured MT data in a certain area of Tongchuan city, Shaanxi province were selected for morphological gradient processing. The results show that the morphological gradient can effectively reduce the random error and achieve the purpose of extracting the target anomaly.