The OECF (Opto Electronic Conversion Function) metric talks about the relevance of the tone transfer, between the scene (our target) and the image, and evaluates if each zone of the image (high lights, half lights and shadows) are in place.
The OECF allows us to make estimates about the relevance of the camera exposure, presence of curves introduced during image processing or to evaluate the neutrality, or balance between the different RGB channels.
Unlike Delta-e evaluation, the OECF is only done on neutral patches and is especially oriented for use with densitometric scales. Since the OECF has to be related to a reference, it is necessary to load a reference document in CGATS format with the colorimetry of the samples to be studied.
The implemented metrics have been expressed as follows:
- OECF: represents the transfer of everything in relation to the reference provided in the reference document. The Y-axis is expressed in the standardized Luma in percentages.
- RGB: The average for each sample is shown graphically for each R, G and B channel so that we can discern the channel balance for each sample
- RED, GREEN and BLUE show the OECF for each channel compared to its reference. The Y axis is expressed in CV or Count Values for 8 bits (0-255 units)
- DEV: expresses the exposure error in terms of EV, i.e. how much error we have made in the exposure (relationship between camera and processing) when generating the image. In an image, where there are no tone curves applied, all samples should have an approximate error, in images with tone curves, high lights and shadows will have disparate errors.
- Err Average: shows the average difference between the values of each sample and its reference.
- Err Dev: is the standard deviation of the average error for each sample
- Err Max and Err Min: are the maximum and minimum errors in the set of patches studied
- Dev Avg, Desv Max and Desv Min: only appear on the RGB panel and indicate the maximum and minimum average deviation for each sample. The higher the average deviation, the more error in neutrality there is in our samples.