Article

Sreelakshmi P M, Federal Institute Of Science And Technology; Paul P Mathai ,Federal Institute of Science And Technology

Taking photos under dim lighting conditions using a hand-held camera is very challenging. If the camera is set to a large exposure time, the image is blurred due to camera shake. On the other hand, if it is taken with a short exposure time, the image is dark and noisy. Various approaches try to remove image blur due to camera shake, either with one or multiple input images. Blur introduced in an image from camera shake is mostly due to the 3D rotation of the camera. This results in a blur kernel which is non-uniform throughout the image. In this method a new algorithm is proposed which do not estimate the blur kernel instead use an align and average like technique. Burst images are considered because each image in the burst is blurred differently. The proposed algorithm performs a weighted average, for the burst images, in the Fourier domain, with weights depending on the Fourier spectrum magnitude. Rationale is that camera shake has a random nature, and therefore, each image in the burst is generally blurred differently. First, the burst of images are registered. Then, for that image correspondences are considered to estimate the dominant homograph relating every image of the burst and a reference image (the first one in the burst).Image correspondences are found using SURF algorithm. Then, Fourier Burst Accumulation is done channel by channel using the same Fourier weights for all channels. Then a noise removal is done using NLMEANS denoising algorithm and finally, a Gaussian sharpening is done on the filtered image. To avoid removing fine details, a percentage of what has been removed during the denoising step is added back finally. Experiments with real camera data show that the proposed burst accumulation algorithm achieves faster and better results.