Method and system for processing an input image based on a guidance image and weights determined thereform

1. A method for processing an input image, the input image comprising a plurality of pixels, the method comprising:

generating a base layer portion and a detail layer portion from the input image by carrying out a linear transformation of pixels of a guidance image;

determining, for each of the pixels of the guidance image, a variance of the pixel in a surrounding window of pixels; and

determining, for each of the pixels of the guidance image, a weight value of the pixel based on the determined variance;

wherein the pixels are weighted in the linear transformation according to the determined weight value, and

wherein the method further comprises:

determining the input image as an initial transmission map of a haze image by using a luminance channel of the haze image, comprising:

estimating an atmospheric light value in a dark intensity channel of the haze image; and

determining the initial transmission map based on the atmospheric light value; and

carrying out a haze removal processing of the haze image based on the initial transmission map.

2. The method of claim 1 , wherein the weight value is determined based on a ratio of a variance of each pixel and the variance of all the pixels in the guidance image.

3. The method of claim 2 , wherein the weight value is larger than 1 if the pixel is an edge pixel, and smaller than 1 if the pixel is a flat pixel.

4. The method of claim 1 , further comprising:

determining the weight values of all the pixels in the guidance image; and

smoothing the weight values with a Gaussian filter.

5. The method of claim 1 , further comprising minimizing a cost function of the linear transformation in carrying out the linear transformation of the guidance image, the cost function based on the guidance image and the input image.

6. The method of claim 5 , wherein the cost function comprises a data term and a regularization term, and where the weight value of the pixel is coupled to the data term to modify the fidelity of the base layer.

7. The method of claim 5 , wherein the cost function comprises a data term and a regularization term, and where the weight value of the pixel is coupled to the regularization term to modify the smoothness of the base layer.

8. The method of claim 1 , further comprising:

determining a value of reconstructed pixels from the base layer portion after carrying out the linear transformation of the pixels; and

averaging the plurality of pixel values of the reconstructed pixels.

9. The method of claim 1 , further comprising carrying out a non-negative sky region compensation during haze removal processing.

10. A system for processing an input image, the input image comprising a plurality of pixels, the system comprising:

a linear transformation module configured to carry out a linear transformation of pixels of a guidance image in generating a base layer portion and a detail layer portion from the input image;

a variance determination module configured to determine, for each of the pixels of the guidance image, a variance of the pixel in a surrounding window of pixels; and

a weightage module configured to determine, for each of the pixels of the guidance image, a weight value of the pixel based on the determined variance;

an input image determination module configured to determine the input image as an initial transmission map of a haze image using the luminance channel of the haze image;

an atmospheric light module configured to estimate an atmospheric light value in a dark intensity channel of the haze image; and

a haze removal module configured to carry out a haze removal processing of the haze image based on the initial transmission map,

wherein the pixels are weighted in the linear transformation according to the determined weight value; and

wherein the initial transmission map is determined based on the atmospheric light value.

11. The system of claim 10 , further comprising a Gaussian filter configured to smooth the weight values of the pixels in the guidance image.

12. The system of claim 10 , further comprising a cost function module configured to minimize a cost function of the linear transformation in carrying out the linear transformation of the guidance image, the cost function based on the guidance image and the input image.

13. The system of claim 10 , further comprising:

a reconstruction module configured to determine a value of reconstructed pixels from the base layer portion after carrying out the linear transformation of the pixels; and

an averaging module configured to average the plurality of pixel values of the reconstructed pixels.

14. The system of claim 10 , further comprising a sky region compensation module configured to carry out a non-negative sky region compensation during haze removal processing.

15. The system of claim 10 , further comprising a detail enhancement module configured to adjust the detail layer portion for better visual quality.

16. A non-transitory computer program readable medium having a program recorded thereon, wherein the program is executed to make a computer process an input image, the input image comprising a plurality of pixels, the computer readable medium comprising:

instructions for generating a base layer portion and a detail layer portion from the input image by carrying out a linear transformation of pixels of a guidance image;

instructions for determining, for each of the pixels of the guidance image, a variance of the pixel in a surrounding window of pixels; and

instructions for determining, for each of the pixels of the guidance image, a weight value of the pixel based on the determined variance;

instructions for determining the input image as an initial transmission map of a haze image using the luminance channel of the haze image;

instructions for estimating an atmospheric light value in a dark intensity channel of the haze image; and

instructions for carrying out a haze removal processing of the haze image based on the initial transmission map;

wherein the pixels are weighted in the linear transformation according to the determined weight value, and

wherein the initial transmission map is determined based on the atmospheric light value.