In this thesis, we are concerned with the image restoration problem which has been formulated in the literature as a system of linear inequalities. With this formulation, the resulting constraint matrix is an unstructured sparse-matrix and even with small size images we end up with huge matrices. So, to solve the restoration problem, we have used the surrogate constraint methods, that can work efficiently for large size problems and are amenable for parallel implementations. Among the surrogate constraint methods, the basic method considers all of the violated constraints in the system and performs a single block projection in each step. On the other hand, parallel method considers a subset of the constraints, and makes simultaneous block projections. Using several partitioning strategies and adopting different communication models we have realized several parallel implementations of the two methods. We have used the hypergraph partitioning based decomposition methods in order to minimize the communication costs while ensuring load balance among the processors. The implementations are evaluated based on the per iteration performance and on the overall performance. Besides, the effects of different partitioning strategies on the speed of convergence are investigated. The experimental results reveal that the proposed parallelization schemes have practical usage in the restoration problem and in many other real-world applications which can be modeled as a system of linear inequalities.