Channel assignment in wireless mesh network (WMN) aims at improving the network throughput by utilizing multiple orthogonal frequency channels to minimize the interference. Interference can be categorized as coordinated and non-coordinated, depending upon the relative location of the interfering links. Compared to coordinated interference, non-coordinated interference has a severe adverse impact on throughput. This thesis is based on the hypothesis that the network throughput can be improved significantly, if channel assignment minimizes non-coordinated interference with priority. We propose a static and centralized channel assignment scheme CCAS to show the effectiveness of the hypothesis. The cluster-based approach of CCAS minimizes non-coordinated interference with reduced complexity. CCAS improves the network throughput by upto 80%, compared to the existing schemes. We propose topology control scheme MATS that constructs low interference multipath network topology using a subset of links from physical topology. We report an additional improvement of upto 10% in the network throughput, when CCAS assigns channels to the links selected by MATS. In the final part of the thesis, we formulate generalized channel assignment as an optimization problem, accounting for real network traffic. The objective of the problem is to select the channels for links such that maximum incident traffic can be transmitted over the links, while ensuring a fair distribution of throughput amongst links and elimination of non-coordinated interference. For a given network and incident traffic, the solution to this problem generates the channel assignment resulting in optimal network throughput. We propose dynamic and distributed scheme LYCAS as an approximate solution to the problem. LYCAS employs MATS to construct network topology and cluster-based approach of CCAS to minimize non-coordinated interference. In addition, it periodically updates the assignment of channels to adapt to the changing traffic load. LYCAS achieves upto 68% of the optimal network throughput and upto 72% of optimal aggregate end-to-end throughput of multi-hop flows. It outperforms the existing schemes by a factor of 2.