Method and apparatus for combining low-dynamic range images to a single image
The invention discloses a method to combine low-dynamic range images with different exposures to construct a high-dynamic range image. The method deals with the motions such that to minimize the chance of such “non-recoverable” occlusions and avoids the necessity of knowing or finding the response functions of the low-dynamic range images. Apart from that, the method is fast enough to be used in a mobile camera device and produce good user experience.
1. A method for combing low-dynamic range images to a single image, the method comprising:
acquiring low-dynamic range images with different EV compensation values;
detecting areas of non-similarity of the images by equalizing image brightness according to mutual 2D histograms;
modifying the images in the non-similarity areas before a Laplacian decomposition to the images;
re-distributing and adapting different fusing weights to the images to fuse a single height dynamic range image.
2. The combing method according to claim 1 , wherein the images exposure and sensitivity settings are captured in a certain sequence: dark-normal-bright-normal-dark-normal-bright or bright-normal-dark-normal-bright-normal-dark.
3. The combing method according to claim 1 , wherein the images are aligned to one of them by equalizing different exposure brightnesses through mutual gammas, calculated from the images histograms and tracking features within the equalized images.
4. The combing method according to claim 3 , wherein the non-similarity areas in the images are found after equalizing the aligned images brightness by mutual gammas, derived from mutual 2D histograms.
5. The combing method according to claim 4 , wherein each non-similarity area is taken from its reference image, converted to the other two exposures, using the mutual gammas, and is copied in some or all of the images of the corresponding exposures.
6. The combing method according to claim 1 , wherein the fusion weights within the non-similarity areas are redistributed among a new local reference image for each area and areas, similar to it, among its neighboring images of different exposures.
7. The combing method according to claim 6 , wherein the weights are adapted after decomposition in a Gaussian pyramid to adjust the global tone mapping and local contrast enhancement.
8. The combing method according to claim 1 , wherein the method further comprising: executing frames alignment to the images while the frames are being captured during a burst capture.
9. The combing method according to claim 8 , wherein the executing frames alignment further comprising:
equalizing brightness of the images via corresponding mutual gamma correction;
detecting features to track in a reference image and tracking the features in each of the other images;
executing homography transforms to warp each image from the other images to match the reference image.
10. The combing method according to claim 1 , wherein the method further comprising:
capturing seven input low resolution images of three exposure and sensitivity settings;
converting a 3 rd and 5 th images via a mutual gammas to pixel brightness values of a 4 th reference image;
combining two difference masks by “OR” logical operation, grouping adjacent non-similar pixels together using connected component labeling to form distinct areas;
analyzing the pixel values in each area to determine which exposure should be used to take each area from.
11. The combing method according to claim 1 , wherein the method further comprising:
replacing the non-similarity areas with brightness-adjusted information from other images, prior to pyramid decomposition.
12. An apparatus for combing low-dynamic range images to a single image, comprising:
an acquisition module used to acquire low-dynamic range images with different EV compensation values;
a detection module used to detect areas of non-similarity of the images by equalizing image brightness according to mutual 2D histograms;
a modification module used to modify the images in the non-similarity areas before a Laplacian decomposition to the images;
a fusing module used to re-distribute and adapt different fusing weights to the images to fuse a single height dynamic range image.
13. An electronic device, comprising:
a processor; and
a memory for storing instructions executable by the processor,
wherein the processor is configured to execute the method of claim 1 .
14. A non-transitory computer-readable storage medium storing executable instructions that, when executed by an electronic device with a touch-sensitive display, cause the electronic device to execute the method of claim 1 .
15. The electronic device of claim 13 , wherein the images exposure and sensitivity settings are captured in a certain sequence: dark-normal-bright-normal-dark-normal-bright or bright-normal-dark-normal-bright-normal-dark.
16. The electronic device of claim 13 , wherein the images are aligned to one of them by equalizing different exposure brightnesses through mutual gammas, calculated from the images histograms and tracking features within the equalized images.
17. The electronic device of claim 16 , wherein the non-similarity areas in the images are found after equalizing the aligned images brightness by mutual gammas, derived from mutual 2D histograms.
18. The electronic device of claim 17 , wherein each non-similarity area is taken from its reference image, converted to the other two exposures, using the mutual gammas, and is copied in some or all of the images of the corresponding exposures.
19. The electronic device of claim 13 , wherein the fusion weights within the non-similarity areas are redistributed among a new local reference image for each area and areas, similar to it, among its neighboring images of different exposures.
20. The electronic device of claim 19 , wherein the weights are adapted after decomposition in a Gaussian pyramid to adjust the global tone mapping and local contrast enhancement.