Greater compression ratio can be achieved while compressing images and videos by using the technique called compressive sensing or compressed sensing (CS). In CS, sparsity of the signal is exploited in order to achieve high compression. CS-based compression converts the images and video frames to a set of $$m<n$$ measurements, which are transmitted to the receiver where the images and video frames are recovered by an efficient reconstruction algorithm with minimum error. Many reconstruction procedures solve least squares problem iteratively to recover the original signal, which ultimately increases the algorithmic complexity and runtime. In this work, split, process and merge technique for image and video reconstruction is proposed, which surpasses the adversities of the iterative algorithms, and provides better reconstruction performance in terms of perception and objective measures. Even with small number of measurements, the peak to signal noise ratio obtained using split and merge technique is above 30 dB, and the structural similarity is above 97% with an accuracy of 98% for both images and videos. The total runtime of the proposed algorithm is around 1.5 s and 9 s for images and videos respectively, which is small compared to iterative algorithms. Thus, the proposed split and merge technique based compressive sensing recovery algorithm proves to perform better in terms of speed and reconstruction quality.