Application for Subjective Auto White Balance Accuracy Measurement
Summanen, Kasperi (2013)
2013
Signaalinkäsittelyn ja tietoliikennetekniikan koulutusohjelma
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2013-05-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201305171137
https://urn.fi/URN:NBN:fi:tty-201305171137
Tiivistelmä
Variation in illumination conditions will create situations where hue of the dominant illuminant is di erent. It can be said that the illumination sources have di erent color temperatures. The human visual system has developed to adapt to these changes of the illumination hue, and thus it is able to maintain the color constancy of the objects. The task of the AutomaticWhite Balance (AWB) is to do the same in digital cameras. The purpose of the AWB is to get the output colors to look as much same as the human eye saw the scene when the image was taken, i.e., as natural.
The main objective of this thesis is to implement a mobile application which can collect data from the accuracy of the AWB algorithm and record the level of the color adaption of eye in di erent illumination conditions. With this application test users can capture images and adjust the colors immediately after the capturing. In this case, the adjustments are done in the same illumination condition as the capturing. This is important because the human visual system will adapt to see white as white in di erent illuminations. The nal image from the camera is transformed manually to look correct in the sRGB color space, that corresponds to the color temperature of 6500K. When capturing images under other illumination sources with di erent color temperature we can measure the di erence with the implemented mobile application even when the AWB algorithm has worked correctly.
The human visual system is not perfect and the color adaption will not be complete in low and high color temperatures. Therefore it is expected to get di erence between the AWB result and the test user preferred output even though the algorithm has performed correctly. In addition, these color adaption results can be used to build behavior model of the color adaption when the adaption is incomplete. Further on, these color appearance models can be used to build better AWB algorithm that imitates the human eye better than before and is not always trying to make
white look as white.
Based on the results of the study this concept is working as thought and designed. Anyhow, it contains few error sources that need to be taken into account when analyzing the collected data. The results from the data analysis can not be used to build an automatic classi er, and thus, human involved evaluation is often needed. This will be a major trend in future studies.
The main objective of this thesis is to implement a mobile application which can collect data from the accuracy of the AWB algorithm and record the level of the color adaption of eye in di erent illumination conditions. With this application test users can capture images and adjust the colors immediately after the capturing. In this case, the adjustments are done in the same illumination condition as the capturing. This is important because the human visual system will adapt to see white as white in di erent illuminations. The nal image from the camera is transformed manually to look correct in the sRGB color space, that corresponds to the color temperature of 6500K. When capturing images under other illumination sources with di erent color temperature we can measure the di erence with the implemented mobile application even when the AWB algorithm has worked correctly.
The human visual system is not perfect and the color adaption will not be complete in low and high color temperatures. Therefore it is expected to get di erence between the AWB result and the test user preferred output even though the algorithm has performed correctly. In addition, these color adaption results can be used to build behavior model of the color adaption when the adaption is incomplete. Further on, these color appearance models can be used to build better AWB algorithm that imitates the human eye better than before and is not always trying to make
white look as white.
Based on the results of the study this concept is working as thought and designed. Anyhow, it contains few error sources that need to be taken into account when analyzing the collected data. The results from the data analysis can not be used to build an automatic classi er, and thus, human involved evaluation is often needed. This will be a major trend in future studies.