Contrast-enhancing seam detection and blending using graph cuts
Thomas Weibel, Christian Daul, Didier Wolf, Ronald Rösch · HAL (Le Centre pour la Communication Scientifique Directe) · 2012
During the image placement onto the compositing surface (mosaic), stitching algorithms try to minimize visual inconsistencies (texture discontinuities), seam induced color gradients, and blurry image regions. These problems are classically processed separately. In this contribution, we describe a two step graph-cut algorithm that combines these issues. In the first step, optimal seam locations are detected while maximizing the contrast of the mosaic. The second step corrects vignetting and exposure differences along the previously determined seams while retaining contrast, hue and saturation of the images. Qualitative and quantitative results demonstrates that the proposed method produces exposure corrected mosaics that are locally sharper than the individual images from the sequence.