Joint auto-exposure tone-mapping system
The present disclosure describes compression of high dynamic range (HDR) images into low dynamic range (LDR) images while saving useful data in the HDR image. HDR images can be formed by merging images with different exposure times into a single image with high bit-depth to capture low light and bright data. LDR images can be generated using a joint auto-exposure and tone mapping system to capture useful details of the HDR input in and LDR output. Therefore, embodiments of the present disclosure create high quality LDR images from HDR images using data from the auto-exposure system of a sensor.
1. A method comprising:
computing a virtual auto-exposure gain based on image statistics;
computing a luminosity probability density function based on the image statistics;
generating a tone-mapping function by computing a scaled luminosity cumulative distribution function based on the virtual auto-exposure gain and the luminosity probability density function; and
converting a high dynamic range (HDR) image to a low dynamic range (LDR) image based on the tone-mapping function.
2. The method of claim 1 , further comprising:
adjusting the virtual auto-exposure gain to obtain a target brightness for the LDR image; and
adjusting an analog gain to obtain the target brightness for the LDR image after adjusting the virtual auto-exposure gain.
3. The method of claim 1 , further comprising:
applying a non-linear function to a luminosity of the HDR image to obtain the luminosity probability density function.
4. The method of claim 1 , wherein computing the scaled luminosity cumulative distribution function comprises:
computing a luminosity cumulative distribution function based on the luminosity probability density function; and
scaling an input of the luminosity cumulative distribution function based on the virtual auto-exposure gain to obtain the scaled luminosity cumulative distribution function.
5. The method of claim 1 , further comprising:
collecting a plurality of exposures using an image sensor; and
generating the HDR image by combining the plurality of exposures.
6. The method of claim 1 , further comprising:
computing a plurality of local tone-mapping functions based on the virtual auto-exposure gain and the luminosity probability density function, wherein each of the plurality of local tone-mapping functions corresponds to an average luminosity value;
computing a local average luminosity value for a region surrounding a pixel of the HDR image; and
selecting the tone-mapping function for the pixel from among the plurality of local tone-mapping functions based on the local average luminosity value and the average luminosity value corresponding to the tone-mapping function.
7. The method of claim 1 , further comprising:
selecting a plurality of sample values from among a plurality of luminosity values in a range determined by a bit width of the HDR image; and
generating a plurality of linear functions located between the plurality of sample values to approximate a continuous tone-mapping function by comparing a value of each linear function with a corresponding value of the continuous tone-mapping function, wherein the tone-mapping function is generated based on the plurality of linear functions.
8. The method of claim 1 , further comprising:
generating a smooth version of the HDR image by denoising the HDR image; and
generating a sharp version of the HDR image by sharpening the smooth version of the HDR image, wherein the tone-mapping function is applied to the sharp version of the HDR image based on the smooth version of the HDR image.
9. The method of claim 8 , wherein:
the HDR image has a higher bit width than the LDR image.
10. The method of claim 1 , further comprising:
computing white balance data for the HDR image, wherein the tone-mapping function is generated based on the white balance data.
11. A method comprising:
collecting a plurality of exposures using an image sensor;
generating a high dynamic range (HDR) image by combining the plurality of exposures;
computing a virtual gain based on image statistics from the image sensor;
adjusting the virtual gain to obtain a target brightness for a low dynamic range (LDR) image;
adjusting an analog gain to obtain the target brightness for the LDR image after adjusting the virtual gain;
computing a luminosity probability density function based on the image statistics;
generating a tone-mapping function based on the virtual gain and the luminosity probability density function; and
converting the HDR image to the LDR image based on the tone-mapping function.
12. An apparatus comprising:
an auto-exposure component configured to compute a virtual auto-exposure gain based on image statistics;
a probability density function component configured to compute a luminosity probability density function based on the image statistics;
a curve generation component configured to generate a tone-mapping function by computing a scaled luminosity cumulative distribution function based on the virtual auto-exposure gain and the luminosity probability density function; and
a tone-mapping component configured to convert a high dynamic range (HDR) image to a low dynamic range (LDR) image based on the tone-mapping function.
13. The apparatus of claim 12 , further comprising:
an analog auto-exposure component configured to adjust an analog gain to obtain a target brightness for the LDR image after adjusting the virtual auto-exposure gain to obtain the target brightness for the LDR image.
14. The apparatus of claim 12 , further comprising:
a component configured to generate a strictly monotonous function.
15. The apparatus of claim 12 , wherein the curve generation component is further configured to:
compute a luminosity cumulative distribution function based on the luminosity probability density function; and
scale an input of the luminosity cumulative distribution function based on the virtual auto-exposure gain to obtain the scaled luminosity cumulative distribution function.
16. The apparatus of claim 12 , further comprising:
an image sensor configured to collect a plurality of exposures using an image sensor.
17. The apparatus of claim 16 , wherein the image sensor further includes:
an image statistics component configured to collect image statistics from the image sensor.
18. The apparatus of claim 12 , wherein the curve generation component is further configured to:
compute a plurality of local tone-mapping functions based on the virtual auto-exposure gain and the luminosity probability density function, wherein each of the plurality of local tone-mapping functions corresponds to an average luminosity value,
compute a local average luminosity value for a region surrounding a pixel of the HDR image; and
select the tone-mapping function for the pixel from among the plurality of local tone-mapping functions based on the local average luminosity value and the average luminosity value corresponding to the tone-mapping function.
19. The apparatus of claim 12 , further comprising:
a noise reduction component configured to generate a smooth version of the HDR image by denoising the HDR image; and
a sharp adder configured to generate a sharp version of the HDR image by sharpening the smooth version of the HDR image, wherein the tone-mapping function is applied to the sharp version of the HDR image based on the smooth version of the HDR image.
20. The apparatus of claim 12 , further comprising:
a white balance component configured to compute white balance data for the HDR image, wherein the tone-mapping function is generated based on the white balance data.