CSAIL Research Abstracts - 2005 link to http://publications.csail.mit.edu/abstracts/abstracts05/index.html link to http://www.csail.mit.edu
bullet Introduction bullet Architecture, Systems
& Networks
bullet Language, Learning,
Vision & Graphics
bullet Physical, Biological
& Social Systems
bullet Theory bullet

horizontal line

Computational Photography

Frédo Durand, Elmar Eisemann, Julie Dorsey, Soonmin Bae & Sylvain Paris


Digital photography has already dramatically changed the ease with which we take and share pictures. We strongly believe that digital photography will also bring a second revolution: The flexibility of digital processing has the potential to greatly improve the quality of pictures. The engineering of traditional cameras, lenses and films has reached impressive results, but has always been limited by the strong constraints of chemistry, optics and analog processes. This led master photographers to create a rich craft and to spend hours to finalize each print, relying on skills that are out of reach of most users. This is even more true for motion pictures, where the sheer number of frames makes the production of quality movies the privilege of a few professionals. With the advent of digital cameras, most of these constraints disappear and the processing possibilities are endless. Powerful image enhancement techniques inspired by physical and phenomenological computer graphics methods can be applied to the data of the sensor in order to generate more compelling pictures.

More information about this research can be found at http://graphics.csail.mit.edu/~fredo/photo.html

Flash Photography Enhancement via Intrinsic Relighting

We have developed a technique that enhances the appearance of photographs shot in dark environments by combining a picture taken with the available light and one taken with the flash. We preserve the ambiance of the original lighting and insert the sharpness and more reliable color information from the flash image. We use the bilateral filter to decompose the two images into detail and large-scale layers. We reconstruct the image using the large scale of the available lighting and the detail of the flash. We detect and correct artifacts due to the flash shadow. Our output images provide the combined advantages of available illumination and flash photography.

Tone mapping
Upper-right: traditional flash photograph. Lower-left: using our new technique, the ambiance of the scene is preserved.
Fast Bilateral Filtering for the Display of High-Dynamic-Range Images

We present a new technique for the display of high-dynamic-range-images, which reduces the contrast while preserving detail. It is based on a two-scale decomposition of the image into a base layer, encoding large-scale variations, and a detail layer. Only the base layer has its contrast reduced, thereby preserving detail. The base layer is obtained using an edge-preserving filter called the bilateral filter. This is a non-linear filter, where the weight of each pixel is computed using a Gaussian in the spatial domain multiplied by an influence function in the intensity domain that decreases the weight of pixels with large intensity differences. We express bilateral filtering in the framework of robust statistics and show how it relates to anisotropic diffusion. We then accelerate bilateral filtering by using a piecewise-linear approximation in the intensity domain and appropriate subsampling. This results in a speedup of two orders of magnitude. The method is fast and requires no parameter setting.

Tone mapping
Computational Darkroom

Black-and-white photography is often associated with striking ``artistic'' pictures, and when digital camera users convert their images to grayscales, they are usually frustrated by the flat look they obtain. While manipulating the contrast and histogram in a photo editing software may help, the outcome is usually disappointing and lacks the punch and quality of fine prints. The goal of our work is to obtain compelling black-and-white pictures inspired by the look of master's photographs. It is clear that some of the discrepancies between fine black-and-white prints and casual snapshots cannot be salvaged after the shot by image processing. This is the case of composition in particular, and in general of what Ansel Adams refers to as image (or optical) management. However, we believe that there is ample room for image-processing enhancement in the tonal management of images. Qualities such as the balance between tonal values, the emphasis or suppression of detail can be edited in the darkroom, or more recently using photo editing software. In fact, advanced digital photographers have developed a new craft and know-how based on photo-editing software, and through complex procedures and patient work, they achieve black-and-white pictures of impressive quality. The required skill and the associated tedium unfortunately make this out of reach of most users. We are building on the craft of photographers to produce compelling black-and-white pictures from digital images. We are developing a tone mapping technique that controls the large-scale tonal balance of an image as well as the strength and variation of detail. Used automatically, this technique will facilitate the manipulation of an input photograph to match the tonal characteristics of example images. For advanced users, it will lso offers interactive controls to rapidly explore the rendition of an image.

Research Support

This work was supported by an NSF CISE Research Infrastructure Award (EIA9802220), the Oxygen consortium, a Deshpande Center grant, an INRIA équipe associée and the MIT-France program.

Part of this research was conducted in collaboration with the Artis team in Grenoble, France.


[1] Elmar Eisemann and Frédo Durand Flash Photography Enhancement Via Intrinsic Relighting In The Proceedings of SIGGRAPH 2004

[2] Frédo Durand and Julie Dorsey. Fast Bilateral Filtering for the Display of High-Dynamic-Range Images. In The Proceedings of SIGGRAPH 2002

horizontal line

MIT logo Computer Science and Artificial Intelligence Laboratory (CSAIL)
The Stata Center, Building 32 - 32 Vassar Street - Cambridge, MA 02139 - USA
tel:+1-617-253-0073 - publications@csail.mit.edu
(Note: On July 1, 2003, the AI Lab and LCS merged to form CSAIL.)