High dynamic range imaging through multi-resolusion spline fusion

High dynamic range (HDR) imaging has become a topic of intense research interest in the fields of computer vision and computer graphics in recent years. Commonly a HDR radiance map of a large range scene is recovered from a set of photos taken with conventional imaging equipment at different exposures. Once a HDR map has been computed, one can apply a tone mapping technique to compress the HDR luminances into a displayable range. Contrastingly we can directly combine images into high quality image using fusion technique. This paper presents an extension to the pyramid approach to fusion with substantially large number of brightness levels. A mask, which controls how much each source image contributes to the synthesis image, is generated (at pixel level) automatically based on the brightness of the source images. A simple procedure drives the image fusion process is defined. The resulting algorithm produces good images and practically does not require manual parameter tuning.