Aim to improve the low resolution and noisy range image from scannerless three-dimensional (3D)LIDAR,a reconstruction approach of sparse range image based on adaptive block grayscale-range Markov random filed (MRF) with optimizing weights is proposed through integrating a monocular camera with high resolution.A grayscale-range MRF multilevel correlogram is established.On this basis a fast interpolation is obtained without the texture copying by using block processing and the resconstruciton speed is improved.The edge penalty factor based on simple linear iterative clustering (SLIC) superpixels segmentation is applied to preserve the image structure details.In order to get a robust performance,both the spatial depth kernel function and grayscale similarity kernel function with adaptive adjustment of standard deviation of kernel funciton for different neighborhood systems are used as guided map.The conjugate gradient algorithm is performed for each neighborhood system to fast optimize the global energy function.The experiments with standard image datasets and real images show that proposed method have better performance than bilinear interpolation,bilateral filter and standard MRF,so that it is effective for realizing the image reconstruction of scannerless 3D LIDAR.