Multi-frame noise reduction method, and terminal

1. A multi-frame noise reduction method, comprising:

using M cameras on a terminal to continuously shoot a same scene for K times, wherein M≥2, K≥2, and each of the M cameras simultaneously collects one frame of image at each time of shooting;

for each time of the shooting in the K times:

registering M frames collected by the M cameras; and

generating a composite frame from the registered M frames based on shift amounts among locations of the M cameras to generate K composite frames;

registering the K composite frames; and

performing temporal and spatial filtering on the registered composite K frames.

2. The method according to claim 1 , wherein the M cameras are in a same plane; and, wherein

the method further comprises:

acquiring a translation quantity between M frames of images shot at each of the K times; and, wherein

registration of the M frames of images is according to the translation quantity.

3. The method according to claim 2 , wherein the M cameras comprise a first camera and a second camera, and the M frames of images shot at each of the K times comprise a first image and a second image respectively corresponding to the first camera and the second camera;

acquiring the translation quantity between the M frames of images shot at each of the K times comprises:

acquiring a translation quantity between the first image and the second image; and registering the M frames of images according to the translation quantity comprises:

selecting the first image as a reference image; and

performing a translation operation on the second image according to the translation quantity to determine a common image region of the first image and the second image.

4. The method according to claim 3 , wherein:

acquiring the translation quantity between the first image and the second image comprises:

obtaining the translation quantity based on L=(f×(t/d))/u, wherein

L is the translation quantity between the first image and the second image;

f is a focus length of the first camera or the second camera;

t is a distance between optical centers of a lens of the first camera and a lens of the second camera;

d is a pixel size of the first image, and the pixel size of the first image is the same as a pixel size of the second image; and

u is a depth of field.

5. The method according to claim 2 , wherein:

generating one frame of composite image according to the registered M frames of images comprises:

performing parallax correction on the registered M frames of images to eliminate a difference in pixel value, between the registered M frames of images at a corresponding pixel, caused by a parallax between the M cameras; and

generating, according to pixel values that are at the corresponding pixel and of the M frames of images obtained after parallax correction, a pixel value that is of the composite image and at the corresponding pixel.

6. The method according to claim 5 , wherein performing parallax correction on the registered M frames of images comprises:

when a difference between pixel values, at a corresponding pixel, of a j th frame of image in the registered M frames of images except a benchmark image and of the benchmark image is greater than a preset threshold, using the pixel value that is of the benchmark image and at the corresponding pixel to replace the pixel value that is of the j th frame of image and at the corresponding pixel; and

when the difference between the pixel values, at the corresponding pixel, of the j th frame of image except the benchmark image in the registered M frames of images and of the benchmark image is not greater than the preset threshold, keeping the pixel value that is of the j th frame of image and at the corresponding pixel unchanged, wherein the benchmark image is a frame of image in the registered M frames of images, and j is valued from 1 to M−1.

7. The method according to claim 1 , wherein:

before performing temporal and spatial filtering on the registered composite K frames, the method further comprises:

performing ghost removal on the registered composite K frames.

8. The method according to claim 1 , wherein K=3.

9. A terminal, comprising:

M cameras, configured to continuously shoot a same scene for K times, wherein M≥2, K≥2, and each of the M cameras simultaneously collects one frame of image at each time of shooting; and

a processor, connected to the M cameras and configured to:

for each time of the shooting in the K times:

register M frames collected by the M cameras; and

generate a composite frame from the registered M frames based on shift amounts among locations of the M cameras to generate K composite frames;

register the K composite frames; and

perform temporal and spatial filtering on the registered composite K frames.

10. The terminal according to claim 9 , wherein the M cameras are in a same plane; and, wherein

the processor is further configured to:

acquire a translation quantity between M frames of images shot at each of the K times; and, wherein

registration of the M frames of images is according to the translation quantity.

11. The terminal according to claim 10 , wherein the M cameras comprise a first camera and a second camera, and the M frames of images that are shot at each of the K times comprise a first image and a second image respectively corresponding to the first camera and the second camera; and, wherein

acquiring a translation quantity between M frames of images that are shot at each time of the K times comprises:

acquiring a translation quantity between the first image and the second image; and

registering the M frames of images according to the translation quantity comprises:

selecting the first image as a reference image; and

performing a translation operation on the second image according to the translation quantity to determine a common image region of the first image and the second image.

12. The terminal according to claim 11 , wherein:

acquiring the translation quantity between the first image and the second image comprises that the processor is further configured to:

obtain the translation quantity based on L=(f×(t/d))/u, wherein

L is the translation quantity between the first image and the second image;

f is a focus length of the first camera or the second camera;

t is a distance between optical centers of a lens of the first camera and a lens of the second camera;

d is a pixel size of the first image, and the pixel size of the first image is the same as a pixel size of the second image; and

u is a depth of field.

13. The terminal according to claim 10 , wherein:

generating one frame of composite image according to the registered M frames of images comprises:

performing parallax correction on the registered M frames of images, so as to eliminate a difference in pixel value, between the registered M frames of images at a corresponding pixel, caused by a parallax between the M cameras; and

generating, according to pixel values that are at the corresponding pixel and of the M frames of images obtained after parallax correction, a pixel value that is of the composite image and at the corresponding pixel.

14. The terminal according to claim 13 , wherein performing parallax correction on the registered M frames of images comprises:

when a difference between pixel values, at a corresponding pixel, of a j th frame of image in the registered M frames of images except a benchmark image and of the benchmark image is greater than a preset threshold, use the pixel value that is of the benchmark image and at the corresponding pixel to replace the pixel value that is of the j th frame of image and at the corresponding pixel; and

when the difference between the pixel values, at the corresponding pixel, of the j th frame of image except the benchmark image in the registered M frames of images and of the benchmark image is not greater than the preset threshold, keep the pixel value that is of the j th frame of image and at the corresponding pixel unchanged, wherein the benchmark image is a frame of image in the registered M frames of images, and j is valued from 1 to M−1.

15. The terminal according to claim 9 , wherein:

before performing temporal and spatial filtering on the registered composite K frames, the processor is further configured to:

perform ghost removal on the registered composite K frames.

16. The terminal according to claim 9 , wherein K=3.