A processing method and an associated device for accessing a spatio-temporal part of a compressed video sequence.The method includes the following steps: obtaining a request for access to the part by identifying a temporal section defined between initial and final times of the initial and final spatial regions—different from the initial region of the sequence at the initial and final times; decomposing the access request into a plurality of elementary requests for access to a video fragment, each elementary request identifying a fixed spatial region to extract in a temporal sub-interval of the temporal section; and exploiting at least one elementary request to access video fragments constituting the part to access.