[Paper Reading] Color Image Processing Pipeline: a general survey of digital still cameras

Paper Site: http://www.cs.toronto.edu/~kyros/courses/2530/papers/Lecture-03/Ramanath2005.pdf

Problem Definition

An overview of the image processing pipeline, first from a signal processing perspective and later from an implementation perspective, along with the tradeoffs involved.

Contribution and Discussion

1. General Pipeline in a DSC

   
   
   Pipeline

2. Pipeline in Detail
   2.1 Preprocessing: The raw data obtained from the sensor needs to be processed before producing a full-color image to remove noise and other artifacts, along with a color processing pipeline to produce an accurate or, alternatively, a pleasing representation of the captured scene. Including linearization, dark current compensation and flare compensation.
   2.2 Whitebalance: The HVS has the ability to map “white” colors to the sensation of white, even though an object has different radiance when it is illuminated with different light sources. To tell the difference of these colors and resotre their ture colors.
   2.3 Demosaicking: Make use of pixel neighborhood information to estimate the values of the pixel colors that were not measured.
   2.4 Color Transformation: Produce the image data from unrendered color space to rendered color space. Rendered color spaces are designed for output purposes and have a limited gamut.
   2.5 Post Processing: Subsequent to all the aforementioned processing, aimed at image appearance enhancement, erasing the objectionable artifacts.
   2.6 Display/Archive: Depending on the reproduction medium (computer monitor, type of printer), the data needs to be further transformed into the appropriate color space.

3. Implementation: Consumers determine what is acceptable and what is not for given price and performance. Not simply science, but human tastes dictate the final steps in the pipeline: a “preferred reproduction” is required, rather than an accurate one. Thus the tradeoff need to be considereed.

4. Future: HDR imaging, a standard for digital camera characterization, larger sensor arrays accompanied by more powerful processors in digital cameras.