Class AutoExposure

Class Documentation

class AutoExposure

Adjusts brightness to between 0 and 1 using a simple auto-exposure method.

Works with 1 or 3 channel images. For 3 channel images it applies autoexposure in luminance space to keep colors consistent.

Public Functions

AutoExposure()

Default constructor using default percentile and update values.

AutoExposure(int update_every)

Constructor specifying update modulo, and using default percentiles.

Parameters:

update_every[in] – update every this number of frames.

AutoExposure(double lo_percentile, double hi_percentile, int update_every)

Constructor specifying low and high percentiles, and update modulo.

Parameters:

  • lo_percentile[in] – low percentile to use for adjustment.

  • hi_percentile[in] – high percentile to use for adjustment.

  • update_every[in] – update every this number of frames.

void update(Eigen::Ref<img_t<float>> image, bool update_state = true)

Scales the image so that contrast is stretched between 0 and 1.

The top percentile is 1 - hi_percentile and the bottom percentile is lo_percentile. Similar to linear ‘contrast-stretch’, i.e. normalization.

Parameters:

  • image[in] – Reference to the image, modified in place.

  • update_state[in] – Update lo/hi percentiles if true.

void update(Eigen::Ref<img_t<double>> image, bool update_state = true)

Scales the image so that contrast is stretched between 0 and 1.

The top percentile is 1 - hi_percentile and the bottom percentile is lo_percentile. Similar to linear ‘contrast-stretch’, i.e. normalization.

Parameters:

  • image[in] – Reference to the image, modified in place.

  • update_state[in] – Update lo/hi percentiles if true.

void update(Eigen::TensorMap<rgb_img_t<float>> image, bool update_state = true)

Apply global RGB auto-exposure in-place stretching constrast between 0 and 1.

Parameters:

  • image[in] – RGB image tensor (H x W x 3), modified in-place.

  • update_state[in] – Update lo/hi luminance percentiles if true.

void update(Eigen::TensorMap<rgb_img_t<double>> image, bool update_state = true)

Apply global RGB auto-exposure in-place stretching constrast between 0 and 1.

Parameters:

  • image[in] – RGB image tensor (H x W x 3), modified in-place.

  • update_state[in] – Update lo/hi luminance percentiles if true.

void update(Eigen::TensorMap<const rgb_img_t<ouster::sdk::core::float16_t>> input, Eigen::TensorMap<rgb_img_t<float>> out, bool update_state = true)

Convert fp16-bit RGB to float and apply global RGB auto-exposure stretching constrast between 0 and 1.

Parameters:

  • input[in] – input RGB image tensor (H x W x 3) as float16_t.

  • out[out] – output RGB image tensor (H x W x 3) as float.

  • update_state[in] – Update lo/hi luminance percentiles if true.