The invention relates to a high-dynamic-range gigapixel video generation method which comprises the steps of: carrying out normal video acquisition by a plurality of local viewing angle cameras, carrying out varied exposure acquisition by a central global camera, carrying out repeated exposure sampling on the global by using different lengths of exposure time, and obtaining details of incident light in different intensity ranges; carrying out exposure fusion on a video image obtained by the central global camera through varied exposure acquisition so as to obtain a global high-dynamic-range low-time-domain-noise video image; carrying out preprocessing on a local video image not subjected to varied exposure acquisition, reducing according to a set scale, matching by utilizing zero-mean normalized cross correlation to find a position of the reduced image in a global graph after varied exposure fusion, cutting out an optimal matching block from the global graph, amplifying the optimal matching block to be the same with the local video image in size, and using the amplified optimal matching block as a reference block; and carrying out feature point extraction matching and transformation on the local video image, and carrying out splicing on the local video image by utilizing a graph cutting technology. By the method, a high-dynamic-range gigapixel video can be obtained.