Image quality metric for HDR images and video
Methods and systems for generating an image quality metric are described. A reference and a test image are first converted to the ITP color space. After calculating difference images ΔI, ΔT, and ΔP, using the color channels of the two images, the difference images are convolved with low pass filters, one for the I channel and one for the chroma channels (I or P). The image quality metric is computed as a function of the sum of squares of filtered ΔI, ΔT, and ΔP values. The chroma low-pass filter is designed to maximize matching the image quality metric with subjective results.
1. A method for generating an image quality metric using color differences, the method comprising:
receiving input mages comprising a reference image and a test mage;
converting the reference image and the test image to an ITP reference image and an ITP test image in the ITP color space;
generating difference images ΔI, ΔT, and ΔP by subtracting corresponding pixel values in the ITP reference image and the ITP test image for the I, T, and P color channels;
filtering the ΔI difference image with an intensity low-pass filter to generate a filtered ΔI image (FΔI);
filtering the ΔT and ΔP difference images with a chroma low-pass filter to generate filtered ΔT and ΔP images (FΔT and FΔP);
generating sums of squared values based on corresponding pixel values of the filtered ΔI, ΔT and ΔP images; and
generating the image quality metric as a function of the sums of squared values.
2. The method of claim 1 , wherein converting either one of the input images to the ITP color space is skipped if the input image is already in the ITP color space.
3. The method of claim 1 , wherein converting either one of the input images from the ICtCp color space to the ITP color space comprises:
copying the I and Cp channels in the ICtCp color space to the I and P channels in the ITP color space; and
generating T channel values in the ITP color space by dividing corresponding Ct channel values in the ICtCp color space by 2.
4. The method of claim 1 , wherein generating the sums of squared values at pixel location (i,j) (s(i,j)) comprises computing
s ( i,j )= FΔI 2 ( i,j )+ FΔT 2 ( i,j )+ FΔP 2 ( i,j ).
5. The method of claim 1 , wherein the function of the sums of squared values comprises one of an average function, a maximum function, a variance function, or a standard deviation function.
6. The method of claim 4 , wherein the image quality metric is generated based on a function of √{square root over (s(i,j))} values.
7. The method of claim 1 , wherein the intensity low-pass filter comprises a separable 2D filter, wherein each 1D filter comprises weights given by
FI=[ 1 5 23 59 80 59 23 5 1]/256.
8. The method of claim 1 , wherein the chroma low-pass filter comprises a separable 2D filter, wherein each 1D filter comprises weights given by
FC=[ 1 1 1 1 1 1 2 2 2 2 3 3 3 4 4 4 5 5 6 6 6 7 7 7 8 8 8 8 8 8 8 8 8 8 8 7 7 7 6 6 6 5 5 4 4 4 3 3 3 2 2 2 2 1 1 1 1 1 1]/256.
9. The of claim 1 , wherein the chroma low-pass filter comprises a frequency response characterized by a piece-wise linear function with a response set to 1 for frequencies below 0.5 cycles per visual degree (cpd), and for frequencies above 0.5 cpd, decreasing from I to 0 at a rate of approximately 0.45 with each doubling of frequency in log10 domain.
10. A method for generating an image quality metric using color differences, the method comprising:
receiving input images comprising a reference image and a test image;
converting the reference image and the test image to an ITP reference image and an ITP test image in the ITP color space;
filtering the intensity components of the ITP reference image and the ITP test image with an intensity low-pass filter to generate a filtered intensity reference image and a filtered intensity test image;
filtering the T and P chroma components of the ITP reference image and the ITP test image with a chroma low-pass filter to generate filtered T and P chroma reference images and filtered T and P chroma test images;
generating difference images FΔI, and FΔT, and FΔP by subtracting corresponding pixel values in the filtered I, and P reference images and the filtered I, T, and P reference test images;
generating sums of squared values based on corresponding pixel values of the FΔI, FΔT, and FΔP images; and
generating the image quality metric as a function of the sums of squared values.
11. The method of claim 10 , wherein the function of the sums of squared values comprises one of an average function, a maximum function, a variance function, or a standard deviation function.
12. The method of claim 10 , wherein the intensity low-pass filter comprises a separable 2D filter, wherein each 1D filter comprises weights given by
FI=[ 1 5 23 59 80 59 23 5 1]/256.
13. The method of claim 10 , wherein the chroma low-pass filter comprises a separable 2D filter, wherein each 1D filter comprises weights given by
FC=[ 1 1 1 1 1 1 2 2 2 2 3 3 3 4 4 4 5 5 6 6 6 7 7 7 8 8 8 8 8 8 8 8 8 8 8 7 7 7 6 6 6 5 5 4 4 4 3 3 3 2 2 2 2 1 1 1 1 1 1]/256.
14. The of claim 10 , wherein the chroma low-pass filter comprises a frequency response characterized by a piece-wise linear function with a response set to 1 for frequencies below 0.5 cycles per visual degree (cpd), and for frequencies above 0.5 cpd, decreasing from 1 to 0 at a rate of approximately 0.45 with each doubling of frequency in log10 domain.
15. A non-transitory computer-readable storage medium having stored thereon computer-executable instructions for executing with one or more processors a method in accordance with claim 1 .
16. An apparatus comprising a processor and configured to perform the method recited in claim 1 .