[libcamera-devel] [PATCH 4/4] ipa: raspberrypi: Add tuning file for IMX296 sensor
Laurent Pinchart
laurent.pinchart at ideasonboard.com
Wed Dec 22 00:39:20 CET 2021
Hi David,
On Mon, Dec 20, 2021 at 08:58:45AM +0000, David Plowman wrote:
> On Mon, 20 Dec 2021 at 07:43, Naushir Patuck wrote:
> > On Sun, 19 Dec 2021 at 23:27, Laurent Pinchart wrote:
> >>
> >> From: Naushir Patuck <naush at raspberrypi.com>
> >>
> >> The Sony IMX296 exists in two versions, colour (Bayer) and monochrome.
> >> The tuning file corresponds to the colour version.
> >>
> >> Signed-off-by: Laurent Pinchart <laurent.pinchart at ideasonboard.com>
> >
> > We might want to do a second pass of tuning at some point, but this is a good
> > starting point.
> >
> > Signed-off-by: Naushir Patuck <naush at raspberrypi.com>
> > Reviewed-by: Naushir Patuck <naush at raspberrypi.com>
> >
> >> ---
> >> src/ipa/raspberrypi/data/imx296.json | 305 +++++++++++++++++++++++++++
> >> src/ipa/raspberrypi/data/meson.build | 1 +
> >> 2 files changed, 306 insertions(+)
> >> create mode 100644 src/ipa/raspberrypi/data/imx296.json
> >>
> >> diff --git a/src/ipa/raspberrypi/data/imx296.json b/src/ipa/raspberrypi/data/imx296.json
> >> new file mode 100644
> >> index 000000000000..2444bd2eb940
> >> --- /dev/null
> >> +++ b/src/ipa/raspberrypi/data/imx296.json
> >> @@ -0,0 +1,305 @@
> >> +{
> >> + "rpi.black_level":
> >> + {
> >> + "black_level": 4096
> >> + },
> >> + "rpi.dpc":
> >> + {
> >> + },
> >> + "rpi.lux":
> >> + {
> >> + "reference_shutter_speed": 19184,
> >> + "reference_gain": 1.0,
> >> + "reference_aperture": 1.0,
> >> + "reference_lux": 432,
> >> + "reference_Y": 13773
> >> + },
> >> + "rpi.noise":
> >> + {
> >> + "reference_constant": 0,
> >> + "reference_slope": 2.957
> >> + },
> >> + "rpi.geq":
> >> + {
> >> + "offset": 185,
> >> + "slope": 0.0105
> >> + },
> >> + "rpi.sdn":
> >> + {
> >> + },
> >> + "rpi.awb":
> >> + {
> >> + "priors":
> >> + [
> >> + {
> >> + "lux": 0, "prior":
> >> + [
> >> + 2000, 1.0, 3000, 0.0, 13000, 0.0
> >> + ]
> >> + },
> >> + {
> >> + "lux": 800, "prior":
> >> + [
> >> + 2000, 0.0, 6000, 2.0, 13000, 2.0
> >> + ]
> >> + },
> >> + {
> >> + "lux": 1500, "prior":
> >> + [
> >> + 2000, 0.0, 4000, 1.0, 6000, 6.0, 6500, 7.0, 7000, 1.0, 13000, 1.0
> >> + ]
> >> + }
> >> + ],
> >> + "modes":
> >> + {
> >> + "auto":
> >> + {
> >> + "lo": 2500,
> >> + "hi": 8000
> >> + },
> >> + "incandescent":
> >> + {
> >> + "lo": 2500,
> >> + "hi": 3000
> >> + },
> >> + "tungsten":
> >> + {
> >> + "lo": 3000,
> >> + "hi": 3500
> >> + },
> >> + "fluorescent":
> >> + {
> >> + "lo": 4000,
> >> + "hi": 4700
> >> + },
> >> + "indoor":
> >> + {
> >> + "lo": 3000,
> >> + "hi": 5000
> >> + },
> >> + "daylight":
> >> + {
> >> + "lo": 5500,
> >> + "hi": 6500
> >> + },
> >> + "cloudy":
> >> + {
> >> + "lo": 7000,
> >> + "hi": 8600
> >> + }
> >> + },
> >> + "bayes": 1,
> >> + "ct_curve":
> >> + [
> >> + 2960.0, 0.5114, 0.3294, 3720.0, 0.4222, 0.4164, 3770.0, 0.4194, 0.4194, 6040.0, 0.2871, 0.5702, 8450.0,
> >> + 0.2285, 0.6454
>
> One thing I've learnt to be a bit careful of is having AWB modes that
> give ranges too far outside the range of the CT curve. The CT curve
> simply gets extrapolated and you can sometimes get some slightly weird
> results. So I'd probably clip all those modes that list 2500K to
> something nearer 2800K or 2900K. But this is a relatively minor thing,
> and actually will afflict some of our other tuning files too. It would
> obviously be a good idea to make the tuning tool handle this, only I'd
> never really appreciated the issues back at that time.
>
> >> + ],
> >> + "sensitivity_r": 1.0,
> >> + "sensitivity_b": 1.0,
> >> + "transverse_pos": 0.01,
> >> + "transverse_neg": 0.01157
> >> + },
> >> + "rpi.agc":
> >> + {
> >> + "metering_modes":
> >> + {
> >> + "centre-weighted":
> >> + {
> >> + "weights":
> >> + [
> >> + 3, 3, 3, 2, 2, 2, 2, 1, 1, 1, 1, 0, 0, 0, 0
> >> + ]
> >> + },
> >> + "spot":
> >> + {
> >> + "weights":
> >> + [
> >> + 2, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
> >> + ]
> >> + },
> >> + "matrix":
> >> + {
> >> + "weights":
> >> + [
> >> + 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1
> >> + ]
> >> + }
> >> + },
> >> + "exposure_modes":
> >> + {
> >> + "normal":
> >> + {
> >> + "shutter":
> >> + [
> >> + 100, 10000, 30000, 60000, 120000
> >> + ],
> >> + "gain":
> >> + [
> >> + 1.0, 2.0, 4.0, 6.0, 6.0
> >> + ]
> >> + },
> >> + "short":
> >> + {
> >> + "shutter":
> >> + [
> >> + 100, 5000, 10000, 20000, 120000
> >> + ],
> >> + "gain":
> >> + [
> >> + 1.0, 2.0, 4.0, 6.0, 6.0
> >> + ]
> >> + }
> >> + },
> >> + "constraint_modes":
> >> + {
> >> + "normal":
> >> + [
> >> + {
> >> + "bound": "LOWER", "q_lo": 0.98, "q_hi": 1.0, "y_target":
> >> + [
> >> + 0, 0.5, 1000, 0.5
> >> + ]
> >> + }
> >> + ],
> >> + "highlight":
> >> + [
> >> + {
> >> + "bound": "LOWER", "q_lo": 0.98, "q_hi": 1.0, "y_target":
> >> + [
> >> + 0, 0.5, 1000, 0.5
> >> + ]
> >> + },
> >> + {
> >> + "bound": "UPPER", "q_lo": 0.98, "q_hi": 1.0, "y_target":
> >> + [
> >> + 0, 0.8, 1000, 0.8
> >> + ]
> >> + }
> >> + ]
> >> + },
> >> + "y_target":
> >> + [
> >> + 0, 0.16, 1000, 0.165, 10000, 0.17
> >> + ]
> >> + },
> >> + "rpi.alsc":
> >> + {
> >> + "omega": 1.3,
> >> + "n_iter": 100,
> >> + "luminance_strength": 0.5,
> >> + "calibrations_Cr":
> >> + [
> >> + {
> >> + "ct": 4000, "table":
> >> + [
> >> + 2.554, 2.554, 2.541, 2.534, 2.495, 2.506, 2.516, 2.517, 2.518, 2.515, 2.513, 2.495, 2.481, 2.533, 2.533, 2.521,
> >> + 2.522, 2.534, 2.539, 2.531, 2.531, 2.506, 2.506, 2.513, 2.513, 2.509, 2.498, 2.496, 2.508, 2.517, 2.521, 2.521,
> >> + 2.509, 2.517, 2.534, 2.529, 2.531, 2.521, 2.517, 2.517, 2.515, 2.514, 2.506, 2.499, 2.508, 2.508, 2.521, 2.537,
> >> + 2.507, 2.508, 2.517, 2.516, 2.495, 2.487, 2.519, 2.534, 2.535, 2.531, 2.499, 2.494, 2.501, 2.511, 2.526, 2.526,
> >> + 2.509, 2.517, 2.507, 2.501, 2.494, 2.519, 2.539, 2.539, 2.537, 2.537, 2.533, 2.499, 2.503, 2.511, 2.529, 2.525,
> >> + 2.521, 2.522, 2.476, 2.501, 2.501, 2.539, 2.546, 2.538, 2.531, 2.538, 2.541, 2.531, 2.529, 2.526, 2.529, 2.525,
> >> + 2.516, 2.519, 2.469, 2.499, 2.499, 2.543, 2.543, 2.531, 2.528, 2.534, 2.541, 2.535, 2.531, 2.526, 2.531, 2.528,
> >> + 2.509, 2.515, 2.465, 2.487, 2.487, 2.539, 2.543, 2.539, 2.533, 2.549, 2.542, 2.531, 2.529, 2.524, 2.532, 2.533,
> >> + 2.499, 2.499, 2.475, 2.482, 2.471, 2.509, 2.539, 2.544, 2.543, 2.545, 2.533, 2.498, 2.521, 2.521, 2.537, 2.536,
> >> + 2.499, 2.488, 2.488, 2.488, 2.471, 2.462, 2.509, 2.539, 2.539, 2.532, 2.498, 2.498, 2.518, 2.518, 2.539, 2.539,
> >> + 2.483, 2.484, 2.488, 2.488, 2.502, 2.496, 2.508, 2.514, 2.518, 2.517, 2.521, 2.518, 2.518, 2.518, 2.525, 2.539,
> >> + 2.483, 2.487, 2.478, 2.478, 2.507, 2.509, 2.514, 2.513, 2.514, 2.517, 2.536, 2.559, 2.501, 2.501, 2.503, 2.525
> >> + ]
> >> + }
> >> + ],
> >> + "calibrations_Cb":
> >> + [
> >> + {
> >> + "ct": 4000, "table":
> >> + [
> >> + 2.619, 2.603, 2.599, 2.597, 2.595, 2.594, 2.589, 2.587, 2.586, 2.589, 2.592, 2.597, 2.601, 2.608, 2.621, 2.621,
> >> + 2.619, 2.615, 2.603, 2.601, 2.596, 2.595, 2.591, 2.589, 2.589, 2.592, 2.599, 2.593, 2.601, 2.613, 2.622, 2.631,
> >> + 2.617, 2.617, 2.612, 2.611, 2.604, 2.598, 2.593, 2.591, 2.592, 2.591, 2.593, 2.595, 2.599, 2.614, 2.623, 2.631,
> >> + 2.624, 2.619, 2.615, 2.612, 2.605, 2.602, 2.597, 2.596, 2.592, 2.592, 2.595, 2.599, 2.602, 2.606, 2.619, 2.624,
> >> + 2.629, 2.627, 2.627, 2.617, 2.609, 2.598, 2.612, 2.623, 2.615, 2.604, 2.589, 2.595, 2.599, 2.608, 2.611, 2.614,
> >> + 2.629, 2.632, 2.637, 2.627, 2.612, 2.612, 2.629, 2.631, 2.628, 2.621, 2.604, 2.597, 2.598, 2.604, 2.609, 2.609,
> >> + 2.635, 2.636, 2.642, 2.628, 2.623, 2.623, 2.636, 2.636, 2.634, 2.628, 2.616, 2.599, 2.597, 2.601, 2.603, 2.601,
> >> + 2.641, 2.639, 2.646, 2.632, 2.627, 2.625, 2.632, 2.635, 2.634, 2.627, 2.614, 2.596, 2.595, 2.599, 2.599, 2.598,
> >> + 2.643, 2.644, 2.651, 2.649, 2.629, 2.617, 2.624, 2.629, 2.625, 2.614, 2.586, 2.599, 2.595, 2.597, 2.592, 2.595,
> >> + 2.645, 2.646, 2.649, 2.649, 2.638, 2.624, 2.616, 2.617, 2.609, 2.604, 2.603, 2.603, 2.595, 2.589, 2.587, 2.592,
> >> + 2.641, 2.643, 2.649, 2.647, 2.638, 2.618, 2.615, 2.608, 2.602, 2.595, 2.596, 2.595, 2.593, 2.584, 2.581, 2.583,
> >> + 2.638, 2.637, 2.647, 2.634, 2.634, 2.618, 2.621, 2.621, 2.611, 2.602, 2.596, 2.583, 2.581, 2.581, 2.576, 2.574
> >> + ]
> >> + }
> >> + ],
>
> I think having colour tables for only one CT is "a bit weak", but
> still considerably better than nothing. I now have my fancy
> super-expensive calibration kit that will let us measure images at
> about 3000K and 6000K, so we should probably use that.
>
> >> + "luminance_lut":
> >> + [
> >> + 1.308, 1.293, 1.228, 1.175, 1.139, 1.108, 1.092, 1.082, 1.082, 1.086, 1.097, 1.114, 1.149, 1.199, 1.279, 1.303,
> >> + 1.293, 1.249, 1.199, 1.162, 1.136, 1.109, 1.087, 1.077, 1.072, 1.081, 1.095, 1.103, 1.133, 1.172, 1.225, 1.282,
> >> + 1.251, 1.212, 1.186, 1.159, 1.129, 1.114, 1.102, 1.088, 1.088, 1.088, 1.095, 1.117, 1.123, 1.158, 1.198, 1.249,
> >> + 1.223, 1.192, 1.177, 1.163, 1.147, 1.139, 1.132, 1.112, 1.111, 1.107, 1.113, 1.118, 1.139, 1.155, 1.186, 1.232,
> >> + 1.207, 1.186, 1.171, 1.162, 1.168, 1.163, 1.153, 1.138, 1.129, 1.128, 1.132, 1.136, 1.149, 1.167, 1.189, 1.216,
> >> + 1.198, 1.186, 1.176, 1.176, 1.177, 1.185, 1.171, 1.157, 1.146, 1.144, 1.146, 1.149, 1.161, 1.181, 1.201, 1.221,
> >> + 1.203, 1.181, 1.176, 1.178, 1.191, 1.189, 1.188, 1.174, 1.159, 1.153, 1.158, 1.161, 1.169, 1.185, 1.211, 1.227,
> >> + 1.211, 1.179, 1.177, 1.187, 1.194, 1.196, 1.194, 1.187, 1.176, 1.169, 1.171, 1.171, 1.175, 1.189, 1.214, 1.226,
> >> + 1.219, 1.182, 1.184, 1.191, 1.195, 1.199, 1.197, 1.194, 1.188, 1.185, 1.179, 1.179, 1.182, 1.194, 1.212, 1.227,
> >> + 1.237, 1.192, 1.194, 1.194, 1.198, 1.199, 1.198, 1.197, 1.196, 1.193, 1.189, 1.189, 1.192, 1.203, 1.214, 1.231,
> >> + 1.282, 1.199, 1.199, 1.197, 1.199, 1.199, 1.192, 1.193, 1.193, 1.194, 1.196, 1.197, 1.206, 1.216, 1.228, 1.244,
> >> + 1.309, 1.236, 1.204, 1.203, 1.202, 1.194, 1.194, 1.188, 1.192, 1.192, 1.199, 1.201, 1.212, 1.221, 1.235, 1.247
>
> That's really a surprisingly small amount of luminance correction!
>
> >> + ],
> >> + "sigma": 0.005,
> >> + "sigma_Cb": 0.005
>
> Measuring two CTs will probably mean it can give us better "sigma" values.
>
> >> + },
> >> + "rpi.contrast":
> >> + {
> >> + "ce_enable": 1,
> >> + "gamma_curve":
> >> + [
> >> + 0, 0, 1024, 5040, 2048, 9338, 3072, 12356, 4096, 15312, 5120, 18051, 6144, 20790, 7168, 23193,
> >> + 8192, 25744, 9216, 27942, 10240, 30035, 11264, 32005, 12288, 33975, 13312, 35815, 14336, 37600, 15360, 39168,
> >> + 16384, 40642, 18432, 43379, 20480, 45749, 22528, 47753, 24576, 49621, 26624, 51253, 28672, 52698, 30720, 53796,
> >> + 32768, 54876, 36864, 57012, 40960, 58656, 45056, 59954, 49152, 61183, 53248, 62355, 57344, 63419, 61440, 64476,
> >> + 65535, 65535
> >> + ]
> >> + },
> >> + "rpi.ccm":
> >> + {
> >> + "ccms":
> >> + [
> >> + {
> >> + "ct": 2960, "ccm":
> >> + [
> >> + 1.61465, -0.49757, -0.11708, -0.30813, 1.53862, -0.23049, 0.04267, -0.76988, 1.72721
> >> + ]
> >> + },
> >> + {
> >> + "ct": 3720, "ccm":
> >> + [
> >> + 1.63316, -0.55061, -0.08255, -0.27238, 1.54441, -0.27203, 0.01948, -0.66461, 1.64513
> >> + ]
> >> + },
> >> + {
> >> + "ct": 3770, "ccm":
> >> + [
> >> + 1.62378, -0.54465, -0.07913, -0.26849, 1.54674, -0.27825, 0.01751, -0.64966, 1.63214
> >> + ]
> >> + },
> >> + {
> >> + "ct": 6000, "ccm":
> >> + [
> >> + 1.56877, -0.55713, -0.01164, -0.13524, 1.48494, -0.34969, 0.01184, -0.59069, 1.57885
> >> + ]
> >> + },
> >> + {
> >> + "ct": 6040, "ccm":
> >> + [
> >> + 1.58628, -0.57451, -0.01177, -0.13959, 1.48858, -0.34898, 0.01138, -0.58747, 1.57609
> >> + ]
> >> + },
> >> + {
> >> + "ct": 8450, "ccm":
> >> + [
> >> + 1.51714, -0.33619, -0.18094, -0.16148, 2.03049, -0.86901, 0.00999, -0.92095, 1.91096
> >> + ]
> >> + }
> >> + ]
> >> + },
> >> + "rpi.sharpen":
> >> + {
>
> Default sharpening tends to come out a bit strong the fewer megapixels
> there are (another thing to look into one day...), so reducing the
> strength here is probably advisable.
>
> So in summary, yes, I think another pass at this would be beneficial
> but it will work "well enough" initially, so:
>
> Reviewed-by: David Plowman <david.plowman at raspberrypi.com>
The IMX296 sensors I have here are the monochrome version, so I can't
really calibrate the colour-related parameters. This tuning file comes
from a branch that was shared by Naush.
Can I merge it as-is and let you play with your fancy calibration kit to
update the tuning data ? :-)
> >> + }
> >> +}
> >> diff --git a/src/ipa/raspberrypi/data/meson.build b/src/ipa/raspberrypi/data/meson.build
> >> index e84cd0990c31..211811cfa915 100644
> >> --- a/src/ipa/raspberrypi/data/meson.build
> >> +++ b/src/ipa/raspberrypi/data/meson.build
> >> @@ -4,6 +4,7 @@ conf_files = files([
> >> 'imx219.json',
> >> 'imx219_noir.json',
> >> 'imx290.json',
> >> + 'imx296.json',
> >> 'imx378.json',
> >> 'imx477.json',
> >> 'imx477_noir.json',
--
Regards,
Laurent Pinchart
More information about the libcamera-devel
mailing list