For delivering multimedia services over (wireless) networks it is important that the mechanisms which negotiate and optimize content delivery under resource constraints take into account user perceived quality in order to improve user satisfaction. Within the scope of the IP Multimedia Subsystem (IMS) architecture a novel application server may be added which handles multi-user multi-flow Quality of Experience (QoE) negotiation and adaptation for heterogeneous user sessions. Based on a mathematical model, which takes into account the characteristics of audio, video and data sessions for QoE optimization, we develop several optimization algorithms to be used by the application server to maximize overall user defined QoE parameters for all ongoing multimedia sessions, subject to network resource constraints. Our results show that a greedy based approach provides a reasonable compromise in terms of run-time and sub-optimality for the overall QoE based resource allocations.