Introduction
This work shows method to increase the effective dynamic range the camera using scene aware coded exposure.
Previous work: Here we see two techniques which generate HDR images using single frame capture.
Spatially varies exposure: The method on the left uses an image sensor with an array of 2×2 pixels tile, which generates 4 images with with different exposure settings. These images are combined to create an HDR image, but the final image resolution is a quarter of the original sensor resolution.
SLM: Another method uses a spatial light modulator to adjust the exposure time of the individual pixels, but this method is not scalable to the higher resolutions because it is complex, bulky and distortive
THIS WORK:
We demonstrate a technique to capture the high dynamic range images using the scene-aware coded-exposure camera. In the coded-exposure camera, the effective exposure time for the individual pixel can be controlled independently.
In this work we use the scene aware algorithm to generate the coding patterns based on the information from the previous frame.
For example, coding matrix for frame n+1 is generated from the image captured in frame n.
Algorithm
- The algorithm starts with initial exposure with a uniform coding matrix
- The output of the sensor is similar to output of a conventional camera because .
- This image is used to generate codes for next frame that decide the exposure time of different pixels.
- The generated code matrix is uploaded to the coded-exposure camera, and the frame is captured.
- The raw encoded image is read out from the camera.
- Then this output is normalized based on the code matrix and exposure time settings that were used to capture this image.
- After normalization, we get the image with a wide dynamic range.
Common technique such as exposure fusion can be used to present the image on conventional displays. - Now, this HDR image is used to refine the masks for the next frame.
Results
