In this paper, we propose a wideband channel estimation scheme for millimeter wave (mmWave) systems, where a reconfigurable intelligent surface (RIS) is deployed to assist the wireless communication between the base station and users. Based on the sparsity nature of the mmWave channel, we transform the channel estimation problem into a sparse signal recovery problem. The common support assumption is no longer satisfied when we consider the wideband effect, therefore, we design adaptive blocks to detect the support of each subcarrier, and propose the adaptive block simultaneous orthogonal matching pursuit (ABSOMP) algorithm. Simulation results demonstrate that the proposed scheme can achieve better performance when compared with conventional compressive sensing (CS) based methods.