Method and unit for generating a radiance map
An embodiment of the invention relates to a method for generating a radiance map in High Dynamic Range (HDR) image creation by calculating mean difference curves for a sequence of images taken with different exposures and a transformation curve from the mean difference curve by an algorithm approximated to the Debevec function, by the means of which a radiance map can be calculated from the taken images. A further embodiment of the invention relates to a unit for performing this method.
1. A method for generating a radiance map, comprising:
calculating mean difference curves f(z) for pixel values z by taking with a camera a sequence of n images with different exposures and calculating for each of n images a mean difference curve comprising mean difference values for each pixel value, and
calculating a transformation curve T(y i ) from the mean difference curves by
iteratively selecting pixel values y and the corresponding difference values from the mean difference curve, wherein starting from the maximum possible pixel value as the first pixel value, the next pixel value is chosen by subtracting the mean difference value f(y i ) for that first pixel value y i from that first pixel value according to the equation
y i+1 =y i −f ( y i ),
where i=0, . . . , t and y 0 =2 t −1 for t bit images until minimum pixel value is achieved,
taking a difference for the chosen pixel values with the powers of 2, so that
T ( y i )= y t−i −2 i , and
transforming each of n exposure images to its corresponding radiance map by subtracting for each pixel of the exposure image the difference of the transformation curve.
2. The method for generating a radiance map according to claim 1 , wherein t=8.
3. The method for generating of a radiance map according to claim 1 , further comprising interpolating the value of T at not calculated pixel values between 0 and the maximum possible pixel value.
4. The method for generating a radiance map according to claim 1 , wherein calculating a mean difference curve f(z) comprises
taking with a camera a sequence of n images of an object with different exposures wherein the object and the camera remain static with respect to each other, wherein the sequence contains one reference image with pixel values 0 to a maximum pixel value and n−1 further images with n≧2, and
calculating for each pixel value of the reference image an array containing x difference values, each array corresponding to one of the n−1 further images, wherein x is the number of times the considered pixel value appears in the reference image.
5. The method for generating a radiance map according to claim 4 , wherein the maximum pixel value is 255.
6. A generation unit for generating a radiance map, comprising:
a first calculation unit configured to calculate mean difference curves f(z) for pixel values z by taking a sequence of n images with different exposures and calculating for each of n images a mean difference curve comprising mean difference values for each pixel value, and
a second calculation unit configured to calculate a transformation curve T(y i ) from the mean difference curves by iteratively selecting pixel values y and the corresponding difference values from the mean difference curve, wherein starting from the maximum possible pixel value as the first pixel value, the next pixel value is chosen by subtracting the mean difference value f(y i ) for that first pixel value y, from that first pixel value according to the equation
y i+1 =y i −f ( y i ),
where i=0, . . . , t and y 0 =2 t −1 for t bit images until minimum pixel value is achieved,
taking a difference for the chosen pixel values with the powers of 2, so that
T ( y i )= y t−i −2 i , and
a transformation unit configured to transform each of n exposure images to its corresponding radiance map by subtracting for each pixel of the exposure image the difference of the transformation curve.
7. The generation unit for generating of a radiance map according to claim 6 , further comprising an interpolation unit configured to interpolate the value of T at not calculated pixel values between 0 and the maximum possible pixel value.
8. The generation unit for generating a radiance map according to claim 6 , wherein the first calculation unit comprises
an imaging unit configured to take a sequence of n images of an object with different exposures wherein the object and the imaging unit remains static with respect to each other and wherein the sequence contains one reference image with pixel values 0 to a maximum pixel value and n−1 further images with n>2,
a first processor configured to calculate for each pixel value of the reference image arrays containing n−1 difference values, each array corresponding to one of the n−1 further images, and
a second processor configured to calculate a mean difference curve for each of the n−1 further images, each mean difference curve representing a mean difference value for each pixel value.
9. A non-transitory computer readable medium including computer program instructions that cause a computer to execute a method according to claim 1 .
10. A computer readable storage medium, comprising a non-transitory computer readable medium according to claim 9 .
11. A method for generation a transformation curve T(y i ), comprising:
taking with a camera a first sequence of m images with different exposures,
calculating a camera response function CRF from the m images by means of the Debevec function, and
calculating n−1 mean difference curves f n−1 (z) for pixel values z by taking a sequence of n images with different exposures and generating a transformation curve T(y i ),
wherein the result of T(y i ) applied to the calculated mean difference curves approximates the camera response function CRF.