Panoramic Image Generation Algorithm Based on Cylindrical Surface Mapping

In real-life scenarios, the field-of-view angles captured by cameras are typically constrained. This has led to a growing demand for panoramic images to cater to various applications, such as virtual reality, mapping, decorative displays, and more. To meet this demand, we proposed an algorithm based on cylindrical mapping for generating panoramic images with a wide field of view angle. The algorithm utilizes a small number of standard cameras to capture images from various angles. It incorporates cylindrical surface mapping into the image-stitching process and continuously expands the view angles by iteratively stitching multiple sets of adjacent images to create a seamless and circular 360-degree panoramic image. The threshold-based SURF algorithm is used for image registration. The normalization processing algorithm is specifically developed to rectify image distortion during stitching. Additionally, the irregular boundary panorama image is transformed into a rectangular shape using variable radius circular surface mapping to eliminate any invalid void regions within the image. The results of the experiment demonstrate that the proposed algorithm successfully achieves panoramic image stitching with good scene adaptability. Compared to existing representative algorithms, it requires fewer camera points and produces wider viewing angle panoramic images.