[libcamera-devel] [RFC PATCH 1/5] libcamera: controls: Add sensor test pattern mode

[Laurent Pinchart]

Hi Hiro,

Thank you for the patch.

On Fri, Apr 09, 2021 at 01:32:04PM +0900, Hirokazu Honda wrote:
> The control is used to report available sensor test pattern modes
> and also specify the mode to sensor.
> 
> Signed-off-by: Hirokazu Honda <hiroh at chromium.org>
> ---
>  src/libcamera/control_ids.yaml | 59 ++++++++++++++++++++++++++++++++++
>  1 file changed, 59 insertions(+)
> 
> diff --git a/src/libcamera/control_ids.yaml b/src/libcamera/control_ids.yaml
> index b4771f9d..be2f710a 100644
> --- a/src/libcamera/control_ids.yaml
> +++ b/src/libcamera/control_ids.yaml
> @@ -608,4 +608,63 @@ controls:
>          detection, additional format conversions etc) count as an additional
>          pipeline stage.
>  
> +  - SensorTestPatternMode:

Does this need to be limited to the sensor, or can we also produce the
test pattern on the ISP side ? I'm thinking about SoCs that have a test
pattern generator in the inline ISP, in the Bayer domain, just after the
CSI-2 receiver.

> +      type: int32_t
> +      draft: true
> +      description: |
> +       Control to select the sensor test pattern mode. Currently identical to
> +       ANDROID_SENSOR_TEST_PATTERN_MODE.
> +
> +        Mode of operation for the test pattern mode.
> +      enum:
> +        - name: SensorTestPatternModeOff
> +          value: 0
> +          description: |
> +            No test pattern mode is used. The camera device returns from the
> +            image sensor.
> +        - name: SensorTestPatternModeSolidColor
> +          value: 1
> +          description: |
> +            Each pixel in [R, G_even, G_odd, B] is replaced by its respective
> +            color channel.

How do we pick the colour ?

> +        - name: SensorTestPatternModeColorBars
> +          value: 2
> +          description: |
> +            All pixel data is replaced with an 8-bar color pattern. The vertical
> +            bars (left-to-right) are as follows; white, yellow, cyan, green,
> +            magenta, red, blue and black. Each bar should take up 1/8 of the
> +            sensor pixel array width. When this is not possible, the bar size
> +            should be rounded down to the nearest integer and the pattern can
> +            repeat on the right side. Each bar's height must always take up the
> +            full sensor pixel array height.
> +        - name: SensorTestPatternModeColorBarsFadeToGray
> +          value: 3
> +          description: |
> +            The test pattern is similar to SensorTestPatternModeColorBars,
> +            except that each bar should start at its specified color at the top
> +            and fade to gray at the bottom. Furthermore each bar is further
> +            subdevided into a left and right half. The left half should have a
> +            smooth gradient, and the right half should have a quantized
> +            gradient. In particular, the right half's should consist of blocks
> +            of the same color for 1/16th active sensor pixel array width. The
> +            least significant bits in the quantized gradient should be copied
> +            from the most significant bits of the smooth gradient. The height of
> +            each bar should always be a multiple of 128. When this is not the
> +            case, the pattern should repeat at the bottom of the image.

Do you know what sensor implements this ? I've never seen it before.

> +        - name: SensorTestPatternModePn9
> +          value: 4
> +          description: |
> +            All pixel data is replaced by a pseudo-random sequence generated
> +            from a PN9 512-bit sequence (typically implemented in hardware with
> +            a linear feedback shift register). The generator should be reset at
> +            the beginning of each frame, and thus each subsequent raw frame with
> +            this test pattern should be exactly the same as the last.

Same question.

> +        - name: SensorTestPatternModeCustom1
> +          value: 5
> +          description: |
> +            The first custom test pattern. All custom patterns that are
> +            available only on this camera device are at least this numeric
> +            value. All of the custom test patterns will be static (that is the
> +            raw image must not vary from frame to frame).

Could we figure out a better way to implement support for custom test
pattern modes ? Having Custom1 immediately following Pn9 won't allow for
new patterns to be defined. Starting at a higher value would be an
option, but I'd like to find a good way to handle custom values, for all
controls that have the same issue.

> +
>  ...

-- 
Regards,

Laurent Pinchart