The rapid growing of wireless multimedia applications increases the needs of spectrum resources, but today's spectrum resources have become more and more scarce and large part of the assigned spectrum is in an inefficiency usage. Cognitive Radio (CR) technologies are proposed to solve current spectrum inefficiency problems and offer users a ubiquitous wireless accessing environment, relying on dynamic spectrum allocation. However, there are two unsolved problems in previous work: 1) based on the simplified Quality of Service (QoS) uniform assumption, specific requirements of different wireless multimedia applications cannot be satisfied; 2) aiming at single-objective optimization of spectrum utilization or handoff rate, the co-optimization of these two necessary objectives in CR networks has not been achieved. In this paper, we propose a Two-tier Cooperative Spectrum Allocation method (TCSA) to solve these two problems. TCSA consists of two functional parts: one is a Spectrum Adjacency Ranking algorithm implemented at the secondary users' terminals to satisfy the QoS requirements for different wireless multimedia applications; and the other is a Max Hyper-weight Matching algorithm implemented at the cognitive engines of CR networks to co-optimize spectrum utilization and secondary users' spectrum handoff rate. Simulation results show that, compared with the other Random matching algorithm and Cost minimized algorithm, TCSA can significantly improve the performance of CR networks in terms of secondary users' throughput and spectrum handoff rate.