The invention discloses a multi-distortion stereo image quality objective evaluation method, and the method comprises two stages: a training stage and a testing stage. The training stage comprises the steps: obtaining a JPEG distortion stereo image, a Gaussian blur distortion stereo image and a Gaussian white noise distortion stereo image of a non-distortion stereo image, respectively constructing three training image sets, and respectively obtaining an image feature dictionary table and an image quality dictionary table, which are different in distortion type, through combined dictionary training, wherein the JPEG distortion stereo image, the Gaussian blur distortion stereo image and the Gaussian white noise distortion stereo image are different in distortion intensity. The testing stage comprises the steps: obtaining the sparse coefficient matrix of each sub-block in a testing stereo image through optimization according to the image feature dictionary tables which are obtained at the training stage and are different in distortion type; calculating an image quality objective evaluation prediction value for testing the stereo image through the sparse coefficient matrixes and the image feature dictionary tables which are obtained at the training stage and are different in distortion type, wherein the image quality objective evaluation prediction value is consistent with a subjective evaluation value better.