The invention relates to a self-focusing method suitable for scene target imaging by a stripmap synthetic aperture sonar (SAS). The method includes the steps of obtaining coarsely-focused image data; carrying out block division in a distance direction and an azimuth direction to obtain sub block units; selecting strong scattering points; carrying out windowing after each strong scatter point is centered through circle moving, and then convoluting with reference LFM signals in the azimuth direction to obtain signal forms of the strong scatter points before azimuthal compression; cutting out a half length of the synthetic aperture from the original position of each selected strong scatter point target to each of the two sides, and carrying out oblique solution processing to obtain oblique solution processing results of the strong scatter points; obtaining gradients of phase errors of the strong scattering points, and summing the gradients strong of the phase errors of the strong scattering points to obtain a corresponding phase error in the azimuth direction; carrying out a compensation by using the phase error in the azimuth direction and then carrying out focus imaging; and determining whether iteration ending conditions are met currently, if not, carrying out the previous steps again, and otherwise, ending the iterative operation to obtain final imaging results.