Owen Carmichael : Research : Super-Resolution

Home / News Affiliations / Contact Research Medical Image Analysis Object Recognition From Images Object Recognition From Range Data 3D Modeling From Range Data Super-Resolution Image Retrieval Papers Teaching Community Resources Reading Groups Leisure

During a summer at Mitsubishi Electric Research Lab (MERL), I helped Bill Freeman in the development of the VISTA method for super-resolution of images. In this framework, a machinery for estimating high-resolution details of blurry images is optimized through the use of training examples of blurry images and their sharp, highly-resolved counterparts. Here’s an example super-resolution result. The image on the left is the low- resolution input, and the image on the right is the higher-resolution output. Note that there are sharp details in the hair and eyelid. Paper: W. T. Freeman, E.C. Pasztor, O. Carmichael. Learning Low-Level Vision. International Journal Of Computer Vision, V. 40 (1), October 2000, pp 25-47. Also appears as Mitsubishi Electric Research Laboratories Technical Report MERL-TR2000-05. [ pdf, 5 MB] Web Sites: Bill Freeman’s web page, Freeman’s page on super-resolution, Freeman’s page on learning low-level vision