In geophysical tomography, accurate determination of medium parameters using crosshole data is a serious problem. This is due to the fact that the distance matrix (D ij ) is highly ill-conditioned, large, sparse, and nearly singular. The matrix equations may either require underdetermined or overdetermined solutions and may become inconsistent due to the presence of system noise. As a result, direct inversion methods invariably fail to give unique solutions. Many image reconstruction techniques have their own limitations and are not governed by the constraints of background data. To reduce nonuniqueness and to improve the quality of the reconstructed image, many workers have suggested that either one should add more diverse data or apply constraints based on preliminary information about the medium.We use a new image reconstruction technique 'RAYPT' (ray-projection technique), similar to algebraic reconstruction technique (ART), to compute average slowness of each ray. Based on average slowness value of the rays, constraints are applied to the rays, helping to accurately estimate the image parameters. By taking rays one by one from the first to the last, the iteration continues until the desired rms error value is obtained. We find that this technique is not only computationally very efficient but also gives very good quality images.