Application of LCLP to lossy source coding

Linear code linear programming (LCLP) proposed by Feldman is one of the promising methods to implement the lossy source coding process efficiently. In the encoding and decoding of a lossy source code constructed by sparse matrices, we propose auxiliary methods to obtain approximate optimal values in the case where the output of LCLP takes non-integer values. In a simulation, the proposed method attains smaller average distortions than the time-sharing bound.