Software Implementation of Control Functions for Automatic Exposure Control (AEC), De-flickering, Color Temperature Measurement and Color Correction

Art der Arbeit:





Masterstudiengang Embedded Systems Design

Zusammenfassung der Arbeit:

Camera sensors plays an important role in the field of intelligent vehicle and active vehicle safety applications. They provide a detailed information about the 360 degree surroundings. For Autonomous Drive (AD), they work as a necessary input parameter for various applications like lane detection, adaptive cruise control, obstacle detection. For all these applications, the Automatic Exposure Control (AEC) is an important library for normalising the image brightness. It is a function which could aid our Image Signal Processing (ISP) pipeline by analysing and modifying the camera sensor signal and applying various mathematical and computational algorithm to get a high quality image. Exposure parameters cannot vary with the changing surrounding luminance, in real time without an AEC function. Thus, the goal of this thesis, is to develop a reusable generic library of AEC, designed to make the brightness of the input image optimized for current scene. To capture a well-exposed image with high SNR ratio, there are several parameters with which exposure can be adjusted: the aperture, the electronic shutter and the amplification gain. In our case, we will be regulating the exposure time and amplification gains: analog, digital and conversion gains. The luminance histogram distribution helps to adjust the exposure parameters, and works as a measurement feature. In order to cover the whole dynamic range of a scene in an image (120d8 - 150d8) typically three exposures are captured. The long exposure with high sensitivity is responsible for capturing dark regions where as short and very short exposures are accountable for bright and very bright regions, respectively. Then, from these three exposures, a merged high dynamic range (HDR) image is computed. There are different technologies to capture the three exposures. For every HDR technology, the exposure for the next frame is computed and translated into the exposure parameters. In response the visual representation of pixels in a form of histogram is being observed and analyzed whether further exposure value needed to be altered. In ordcr to avoid swing-overs 1" order IIR filter is been implemented. One of the important function part of our AEC Library is the De-flickering function. Flickering is the periodic variation in some light sources and this artifact is been generated due to the electronic rolling shutter technology used in various camera sensors. Some rows of our pixel data integrate during a relatively darker interval of ambient light frequency and some integrate during a brighter interval. Correlation is been used to indicate the frequency. The pixel data get interfered out of the two predominant frequencies(50H2 or 60Hz). For de-flickering the exposure time should be equal to n-integer multiple of ambient line frequency. Another important topic that has been covered in this thesis report is Color Correction Matrices. Color Correction is the processing step to replicate the true colors of the scene. For Color Correction to be implemented it is important to measure the color temperature of the illuminant. Different kinds of illumination of the same scene gives different colour temperature perception, adequately to the illumination. Color temperature is derived from a relationship between temperature of black body and its appeared color. The Maccy's approximation algorithm is been implemented to find the color temperature. In the algorithm, the RGB gain factors from already developed white balancing function (implemented as a post-processing step. The aim of this is to basically want all white objects in the scene to be white in the image.) transformed into XYZ Tristimulus co-ordinates, where then color temperatures are computed using an empirical formula derived using CIE 1931 x, y chromaticity diagram. For correction matrices the CCM tables at seed temperatures are measured using IMATEST tool (an Image Quality Testing Software). For color temperatures lying in between the calibrated seeds, the resulting CCM, interpolated from the nearest calibrated color temperature. Then, color corrected output R,G,B gain values are computed by performing matrix multiplication of color correction matrix and uncorrected R,G,B gain values.