A fundamental problem in digital photography is the estimation of scene colorimetry from raw DSC image data. Currently, a standard is under development in this area (ISO 17321-2). In the development of this standard, few subjective experiments have been carried out until now relating to the estimation of scene colorimetry from noncolorimetric raw data, where no assumptions are made concerning the spectral radiance correlation statistics of the scene. Furthermore, there is not much information available concerning whether it is appropriate to use some assumption about spectral radiance correlation statistics when the statistics of the actual natural scene are unknown.This paper presents the first part of a study involving psychophysical tests to answer the following questions that are essential for the specification of a scene analysis color space, and for the specification of methods for transforming raw DSC data into scene colorimetric data.1. What is the most appropriate error metric to be used for the determination of transformations from raw DSC data to scene colorimetry estimates, when no assumptions are made concerning the scene spectral radiance correlation statistics? The crucial point is to find the error metric that corresponds best with human perception.2. How does this new error metric compare to existing criteria, and how do the existing criteria compare to each other when used to determine transformations based on specified spectral radiance correlation statistics assumptions?3. Given that optimal error metrics are used to determine transformations, how do several spectral radiance correlation statistics assumptions compare to each other and to the maximum ignorance case when applied to natural scenes where the actual statistics are unknown?Several of the transformation methods specified in ISO 17321, and other methods that are extensions of the 17321 methods were applied to raw data from two DSCs with different spectral sensitivity characteristics. These DSCs were used to capture images of a variety of natural scenes, and the resulting images were processed using the different characterization transforms based on different error metrics. Critical visual evaluation of the final images by expert observers was used to eliminate the obviously poor characterizations. Psychophysical experiments were conducted to differentiate the performance of the remaining candidates.