The imaging process can be modified to extract richer information regarding the scene of interest. This modification includes the sensor and complementary algorithms. In this talk, we will focus on two examples of the above approach. First, we show a method to easily remove the effects of haze from scenes and derive information about the scene structure. The method uses an algorithm to extract information based on polarization modulated images. A different algorithm enables the separation of reflections from transparent objects in general scenes, based on the same sensing process.

As our second example, we present the concept of "generalized mosaicing". Here, spatially varying filters are incorporated into a video imaging system. Algorithms applied to sequences taken with such a system can derive high-dynamic range images, multispectral images, very wide depth of field, and other scene information.

This approach has implications for a wide range of imaging modalities. We expect our results to find direct applications in computational vision, computer graphics, remote sensing, astronomical imaging, and medical imaging.