Hyperspectral image analysis mainly concentrates on handling big satellite datasets efficiently and to identify the endmembers more accurately. This paper proposes improved pure pixel identification algorithms to identify the endmembers in hyperspectral images. In the proposed endmember extraction algorithms the skewers are generated based on the statistical parameters of the hyperspectral dataset, which implicitly changes the vertices of the convex hull. This reduces the false alarm probability of the conventional Pixel Purity Index algorithm. Moreover in the proposed Skewer based NFINDR algorithm eliminates the random initialization of the endmembers in the first step, which leads to more promising results. The running time is reduced by decreasing the floating point operations involved. Experimental results validate the effectiveness of the proposed endmember extraction algorithms in terms of improved accuracy and reduced computational complexity. My study proves the proposed algorithms were able to identify the endmembers accurately even in a noisy environment, thereby validating its effectiveness.