introduction to tuning work

The tuning work flow would be the following:

Before we started

Make sure the surface of sensors and lens clean and clear.
Get the prime lens focus by manual.
Turn off 3A fucntion, set gain to 1x and take pictures with appropriate expourse time which means not to be overexposure or underexposure.

Get the sensor under darkness, then take pictures in rawdata format.
Statistics about individual values of four components in the bayer RGGB (red, green1, green2, blue) indicate the offset value which should be cut off to reduce the dark current effect.

We would like to place the color checker in a color light box, and take pictures under four color temperatures (D65, D50, CWF and A Light).
Tunning tools then utilize the raw picture data, and output a group of parameters which are color correction martixs regard to four color temperatures.
ISP should be able to calculate out the correct gain value of red, green1, green2 and blue for color constancy according to the parameters.

The camera would be illuminated with a uniform light source by a light panel. Then pictures get captured.
Tunning tools generate the deshading matrix with the rawdata.

The camera would be illuminated by the TE-251 chart with D65 light source. Then pictures get captured.
Like LSH calibration, tunning tools generate parameters with the rawdata.

Adjusting parameters below in tunning tools.

DNS_STRENGTH

Denoising strength.
DNS_GREYSC_PIXTHRESH_MULT

Pixel threshold multiplier used to adjust the sensitivity of colour differences for the greyscale weights.

The scene for tuning sharpness would typically be a lab scene with charts and real objects. A chart like ISO12233 is a good choice and real objects like dolls, wool (items that have fine texture).
Change the sharpening radius in 0.5 increments until it gets to 2.5 and determine optimal value.
Change sharpening strength in 0.1 increment until reasonable results at edges are achieved.
Change sharpening details in 0.1 decrement (if necessary) to reduce noise in flat regions of the image without removing the sharpening at edges.
After tuning these values it should be validated over a number of scenes/light levels to understand if further fine tuning is required.
The secondary denoiser has got two controls and assuming that sharpening has been done before the process would be as below:
- Increase the secondary denoiser strengh to increase noise reduction.
- Increase the edge avoidance to perform more edge avoidance.

Tone Mapping is to map one set of colors to another to approximate the appearance of high-dynamic-range images in a medium that has a more limited dynamic range.
The automatic tone mapping control (TNMC)is in charge of:
- Dynamically generate a global mapping curve based on previous captures’ statistics.
- Control the strength of the tone mapped module so tone mapping strength is reduced when the sensor is setup in a way which may produce noisy images (basically when Sensor’s gain value is high).
We would like to modify the default gamma curve in tunning tools to generate a new curve (TNMC works based on it) for desired brightness, contrast and saturation.