This paper presents a novel approach to creating full-view panoramic mosaics from image sequences, which does not require controlled camera motions or specific constraints on image acquisition. Unlike traditional methods that often rely on pure horizontal camera panning, this system can handle uncontrolled 3D camera rotations, making it suitable for images taken with handheld digital cameras. The authors use a set of transformations to represent the image mosaics, avoiding singularities common in cylindrical or spherical maps. They introduce a fast and robust algorithm that recovers 3D rotations using a 3-parameter rotational motion model, compared to the 8-parameter planar perspective transform used in previous methods. The paper also includes techniques for estimating camera focal lengths and closing gaps in panoramic sequences due to registration errors. Additionally, the authors propose an efficient method for extracting environment maps from the mosaics, allowing for seamless integration into 3D graphics viewers and hardware. The paper concludes with a discussion on the advantages of their approach and future work, including the development of a dephasing technique to handle significant motion parallax.This paper presents a novel approach to creating full-view panoramic mosaics from image sequences, which does not require controlled camera motions or specific constraints on image acquisition. Unlike traditional methods that often rely on pure horizontal camera panning, this system can handle uncontrolled 3D camera rotations, making it suitable for images taken with handheld digital cameras. The authors use a set of transformations to represent the image mosaics, avoiding singularities common in cylindrical or spherical maps. They introduce a fast and robust algorithm that recovers 3D rotations using a 3-parameter rotational motion model, compared to the 8-parameter planar perspective transform used in previous methods. The paper also includes techniques for estimating camera focal lengths and closing gaps in panoramic sequences due to registration errors. Additionally, the authors propose an efficient method for extracting environment maps from the mosaics, allowing for seamless integration into 3D graphics viewers and hardware. The paper concludes with a discussion on the advantages of their approach and future work, including the development of a dephasing technique to handle significant motion parallax.