The invention discloses an aerial image hybrid segmentation algorithm based on a novel Markov random field and region merging, and the algorithm comprises the following steps: S1, reading a to-be-segmented color image, and converting the to-be-segmented color image into a gray image; s2, performing coarse segmentation on the grayscale image; s3, performing multi-value connected domain informationstatistics; s4, according to the multi-value connected domain information, carrying out region smoothing on the grayscale image after coarse segmentation; and S5, according to the image after region smoothing, carrying out region merging. According to the invention, a novel Markov model with a variable unit of a regional level, regional edge information fused into a potential function and an iterative stop criterion is adopted to smooth a coarsely segmented image; the method can effectively improve the updating rate of the region label, adaptively controls the updating of the region label, andeffectively reduces the 'over-segmentation rate' and 'wrong segmentation rate' during the segmentation of an image with multi-spot noise, higher intra-domain heterogeneity and difficult gradient information extraction.