In hyperspectral image processing, spectral unmixing is a key computation for extracting endmembers and determining their fractional abundances from mixed pixels. Despite the fact that the sparse unmixing model has recently attracted a lot of attention, the acquired hyperspectral images are influenced by noise, indicating that the sparse unmixing model needs to be improved in order to extract the best collection of endmembers. The Lion-Rider algorithm is proposed in conjunction with a linear sparse spectral unmixing to address this complexity. The proposed optimization algorithm determines the fractional abundances of the endmembers bydefining a multi-objective function. The objective function includes Euclidean and Frobenius norms and optimization variables that are associated with the abundance and mixing matrices. The proposed lion-rider optimization approach is the modified rider optimization algorithm whichuses lion optimization algorithm as anintegral component of computation.To define the objective function in estimating fractional abundances, the optimization variables namely Reconstruction Error, Sparsity, Spatial Neighbor, and Spatial Neighbor Correlation are used simultaneously. Urban and Cuprite hyperspectral image datasets are used to evaluate the proposed method. Based on the performance analysis, the proposed lion-rider optimization approach outperformed its competitors namely,Bilinear and Trilinear MultiObjective Spectral Unmixing, Robust Collaborative Nonnegative Matrix Factorization, Pareto-MultiObjective Spectral Unmixing, and Rider Optimization, in estimating the endmembers and their correspoinding fractional abundances with a minimal reconstruction error and root mean square error of 22.4027 and 0. 0297, respectively.