Sufficient Simple in-Camera Imaging Processing Pipeline for RAW-Images and sRGB Images

In this paper, we focus on the problem of modeling in-camera imaging processing pipeline. The key idea of the proposed pipeline is to combine several consecutive modules into only two-stage modules for each camera setting, namely, linear transformation module and non-linear transformation module. The major benefit of the pipeline is simple to reproduce the color such that the camera can obtain consistent robust result across changes in the scene illumination changes or camera-to-camera transfer in any image processing algorithms. To evaluate the performance of the proposed pipeline, the pipeline is verified with two different use cases, i.e. (1) reconstruction of RAW image from sRGB image compared to the ground-truth RAW image with various camera brands/models and different camera parameter settings, and (2) camera-to-camera transfer compared to the ground-truth and accommodating Photoshop image by expert (classical method). The quantitative results in term of RMSEs are mostly less than 3% (e.g., 0.0075 out of the intensity range [0, 1]) which is insignificantly different when comparing with the conventional method. This pipeline presents a simpler than the classical and conventional pipeline. Not only simple pipeline but also less computation is presented.