Abstract: It is a crucial problem for crop disease diagnosing to effectively extract disease spots from crop pictures with complex background. To solve this problem, a new multiple domain multiple phase level set method is proposed. First, the spatial distribution characteristic that the disease spots to be segmented are statistically located in inner parts of crop leaves is adopted to build a multiple phase level set model with inner ordering. Then, to enhance segmentation effectivity for complex background cases, we adopt multiple spatial domains to the model for the first time so that inner and outer level set functions will evolve in different spatial domains. Experimental results show that this method can well extract disease spots for complex background cases with an accuracy rate of 93.3% and will be a good base for diagnoses.