Rapid developments in information and communication technology in recent years have posed a significant challenge in wireless multimedia sensor networks (WMSNs). End-to-end delay and reliability are the critical issues in multimedia applications of sensor networks. In this paper we provide a new cross-layer approach for provisioning the end-to-end delay of the network at a desirable level of the packet delivery ratio (PDR), used here as a measure of network reliability. In the proposed multi-level cross-layer (MLCL) protocol, the number of hops away from the sink is used to set a level for each node. A packet is routed through the path with the minimum hop count to the sink using this level setting. The proposed protocol uses cross-layer properties between the network and medium access control (MAC) layers to estimate the minimum delay, with which a node can deliver a packet to the sink. When a node wants to send a packet, the MLCL protocol compares this minimum delay with the time to live (TTL) of a packet. If the TTL of the packet is higher than the minimum delay, the node sends the packet through the path with the minimum delay; otherwise, the node drops the packet as the node cannot deliver it to the sink within the TTL duration. This packet dropping improves network performance because the node can send a useful packet instead of an unusable packet. The results show a superior performance in terms of end-to-end delay and reliability for the proposed protocol compared to state-of-the-art protocols.