An improved C-V method was proposed aiming at unfavorable result and low efficiency when the original C-V method was used to segment non-binary image. Considering that the original C-V method only used the gray information, the improved C-V method added local variance information of the image. In addition, the improved C-V method set a weight parameter k, and by adjusting k it can control the proportions of the driving force based on the gray and variance information in the process of segmenting the entire image. Thus, this method can segment non-binary image by the regional gray and variance information. Simultaneously, the implicit scheme was used in the process of the numerical realization of the improved method. The experimental results show that the improved C-V method can extract edge of non-binary image more accurately and reduce the times of iteration.