Blooming filter for multiple exposure high dynamic range image sensors
Multiple-exposure high dynamic range image processing may be performed that filters pixel values that are distorted by blooming from nearby saturated pixels. Pixel values that are near saturated pixels may be identified as pixels that may be affected by blooming. The contributions from those pixels may be minimized when producing a final image. Multiple-exposure images may be linearly combined to produce a final high dynamic range image. Pixel values that may be distorted by blooming may be given less weight in the linear combination.
1. A method for processing images, comprising:
at an image processor, receiving a plurality of input images having associated pixel values and associated dynamic ranges;
identifying pixel values in the plurality of input images that are distorted by nearby high pixel values;
combining the plurality of input images to form an output image while minimizing a contribution of the distorted pixel values, wherein the output image has an associated dynamic range image that is higher than the associated dynamic ranges of the input images, wherein identifying the pixel values in the plurality of input images that are distorted by nearby high pixel values comprises, for each pixel value, identifying an associated maximal regional pixel value in a region surrounding the pixel value.
2. The method defined in claim 1 , wherein combining the plurality of input images to form an output image while minimizing the contribution from the distorted pixel values comprises linearly combining the plurality of input images while minimizing the contribution from the distorted pixel values.
3. The method defined in claim 1 , wherein linearly combining the plurality of input images while minimizing the contribution from the distorted pixel values comprises:
for the given pixel position, linearly combining pixel values in the plurality of input images that correspond to the given pixel position to produce a pixel value in the output image that corresponds to the given pixel position, wherein linearly combining pixel values comprises weighting each pixel value using the associated maximal regional pixel value for the pixel.
4. The method defined in claim 1 , wherein identifying the pixel values in the plurality of input images that are distorted by nearby high pixel values comprises identifying pixels values that are distorted by blooming charge effects.
5. The method defined in claim 1 , further comprising:
identifying saturated pixel values in the plurality of input images, wherein combining the plurality of input images to form an output image comprises minimizing a contribution from the saturated pixel values.
6. A method for processing images, comprising:
at an image processor, receiving a plurality of input images having associated pixel values and associated dynamic ranges, wherein the plurality of input images have different associated exposure times;
identifying pixel values in the plurality of input images that are distorted by nearby high pixel values; and
combining the plurality of input images to form an output image while minimizing a contribution from the distorted pixel values, comprising:
for each pixel position, weighting pixel values in the plurality of input images that correspond to the given pixel position, wherein the distorted pixel values are given less weight; and
for each pixel position, combining the weighted pixel values to produce a pixel value in the output image that corresponds to the given pixel position, wherein the output image has an associated dynamic range image that is higher than the associated dynamic ranges of the input images, wherein identifying the pixel values in the plurality of input images that are distorted by nearby high pixel values comprises, for each pixel value, identifying an associated local maximum pixel value from a window surrounding the pixel value, wherein pixel values that have high associated local maximum pixel values are considered to be distorted.
7. The method defined in claim 6 , wherein weighting the pixel values in the plurality of input images that correspond to the given pixel position comprises weighting pixel values with the associated local maximum pixel values.
8. The method defined in claim 6 , wherein weighting the pixel values in the plurality of input images that correspond to the given pixel position comprises weighting the pixel values with the associated local maximum pixel values and at least one threshold value.
9. The method defined in claim 6 , wherein weighting the pixel values in the plurality of input images that correspond to the given pixel position comprises weighting the pixel values with a difference between a threshold value and a pixel value at the given pixel for a given one of the plurality of input images.
10. The method defined in claim 6 , wherein weighting the pixel values in the plurality of input images that correspond to the given pixel position comprises weighting the pixel values with the associated local maximum pixel values and at least two threshold values.
11. The method defined in claim, 6 , wherein weighting the pixel values in the plurality of input images that correspond to the given pixel position comprises:
for each pixel position, using exclusively a pixel value in the input image having an associated exposure time that is greater than the associated exposure times of the other input images when the associated local maximum pixel value of the pixel value is less than a first threshold value.
12. The method defined in claim 6 , wherein weighting the pixel values in the plurality of input images that correspond to the given pixel position comprises:
for each pixel position, using exclusively a pixel value in the input image having an associated exposure time that is less than the associated exposure times of the other input images when the associated local maximum pixel value of the pixel value is greater than a second threshold value.