The work aims to solve these defects, such as low anti-plaintext attack ability induced because the pixel scrambling and diffusion of the current image algorithm are not related to the plaintext, and the poor security caused by periodicity in the scrambling process. The image encryption algorithm based on chaotic segmentation projection strategy and gravity model was proposed. Firstly, the overlapping sub-blocks were obtained by chaotic segmentation of the input plaintext, and the calculation model for the position coordinates of each sub-block was built to output the positions of image sub-blocks. Then, the initial value of Kent mapping was calculated according to the quantity of plaintext pixels, and the random sequence set was obtained by iterative Kent mapping. Overlapping sub-blocks were confused by defining the position scrambling mechanism. The pixel positions in each scrambled sub-block were interchanged by introducing the segmentation project strategy and the scrambled cipher was outputted by reorganizing theses sub-blocks. Finally, the gravity model was improved by designing the quality change function of the pixel to substitute the fixed quality for pixel encryption of the scrambled cipher. The experimental results showed that, compared to the current image encryption technique, the proposed algorithm had greater security, and the correlation of two adjacent pixels in the outputted cipher was the minimum, with stronger anti-plaintext and anti-shear attack performances. The proposed encryption algorithm can be better used for the safe transmission of image information and has better application value.