Camera denoising and sharpening parameters are device related rigid parameters which are programmed in phone camera device. The current tuning method depends solely on manual modulation and visual evaluation of image quality, which is time consuming and difficult to optimally achieve. To this end, we will introduce an automatic tuning method for mobile cameras in this paper, which can tune the WNR parameters automatically and produce high quality images within a feasible processing time. The method contains two parts, a perception model and an optimization algorithm. For the first part, we developed a perception model to evaluate the image quality for mobile cameras through modified CPIQ metrics. For the second part, in order to overcome a high-dimension non-convex optimization problem, we developed a searching strategy to find the optimal solution by conducting quantization and iteratively minimizing the error metric of the perception model.
The IEEE P1858 CPIQ Standard is a new industry standard for assessing camera image quality on mobile devices. The CPIQ standard provides test methodologies for evaluating seven image attributes: spatial frequency response, texture blur, visual noise, color uniformity, chroma level, lateral chromatic displacement, and local geometric distortion. In addition, the CPIQ standard provides mathematical transforms between objective metric values and perceived image quality quantifiable in just noticeable differences, and a framework to combine individual attributes into prediction of overall image quality. This study aims at validating the CPIQ set of image quality metrics and the CPIQ prediction of overall image quality. The two key components of the study are objective measurements of image quality in the lab and subjective evaluation of real-world images by human observers. Nine smartphones were used in the study, with the expected camera quality ranging from low to high. The CPIQ methodology was implemented and practiced in an industrial lab, and measurements of the CPIQ metrics were obtained at varying lighting conditions. The subjective evaluation study was performed in a university lab, using paired comparison and softcopy quality ruler as test methods. The results from this study revealed that objective measurements defined in the CPIQ standard are highly correlated with perceived image quality.