Granted Patent B2

US 12,530,753 · App. 18/475,568 · Granted Jan 20, 2026

Inventors: Srenivas Varadarajan (Karnataka, IN); Anamika Sharma (Karnataka, IN); Kaustav Chanda (Karnataka, IN)

Assignee: SAMSUNG ELECTRONICS CO., LTD.

G06T5/90G06T7/0002G06T7/11G06T7/13G06T7/20G06T2207/20208G06T2207/30168

Abstract

Claims (75)

1 . A method for estimating a perceptual quality of a high dynamic range (HDR) image by an electronic device, the method comprising:

dividing an image into at least one static region and at least one dynamic region;

determining at least one preserved edge in the at least one static region based on a static HDR quality index (SQI);

determining an amount of ghosting in the at least one dynamic region based on a dynamic HDR quality index (DQI);

weighting at least one artifact in the at least one static region and the at least one dynamic region through a visual saliency map;

computing at least one edge preservation score based on the SQI and the DQI;

generating an output HDR quality metric by modulating the at least one computed edge preservation score on an input HDR quality metric; and

estimating a perceptual quality of the HDR image based on the generated output HDR quality metric.

2 . The method as claimed in claim 1 , further comprising applying the generated output HDR quality metric as feedback for tuning at least one parameter to optimize the perceptual quality of the HDR image.

3 . The method as claimed in claim 1 , wherein the SQI determines an HDR blur through at least one lost LDR edge in the at least one static region of the HDR image due to at least one of an overexposure or an underexposure.

4 . The method as claimed in claim 1 , wherein the DQI determines an HDR ghosting through additional edges in the at least one dynamic region of the HDR image due to a foreground motion.

5 . The method as claimed in claim 1 ,

wherein the visual saliency map is used to weigh an edge-cost in the HDR image compared to at least one LDR image before computing the edge preservation score,

wherein the edge-cost corresponds to edge loss and a presence of extra edges in the HDR image.

6 . The method as claimed in claim 1 , wherein the SQI is generated by:

accumulating all edges present in the at least one static region of an input LDR sequence into one edge map;

detecting a blur in an output HDR image by computing a number of edge pixels in the accumulated edges that are not found in the output HDR image at a corresponding location;

computing the blur as a fraction of edge pixels in the accumulated edge map that are not present in the output HDR image; and

subtracting the computed blur ratio from a predefined value to obtain the SQI, wherein the predefined value is 1.

7 . The method as claimed in claim 1 , wherein the DQI is generated by:

counting a number of edge pixels in the at least one dynamic region for each of LDR images in the input sequence;

determining the LDR image containing a maximum number of edges;

accumulating the edges from the LDR image and the edges in the output HDR image;

detecting the at least one artifact as extra edges present in the output HDR image that are not present in the input LDR image that has the maximum number of edges; and

computing the DQI as the maximum number of edge pixels in an input sequence divided by the number of edges in the accumulated edge map, wherein the computation is always less than or equal to one, where the computation is equal to one if there are no extra edge pixels in the output HDR image.

8 . An electronic device, comprising:

a processor;

a memory; and

a perceptual quality and estimation controller, coupled with the processor and the memory,

wherein the perceptual quality and estimation controller is configured to:

divide an image into at least one static region and at least one dynamic region;

determine at least one preserved edge in the at least one static region based on a static high dynamic range (HDR) quality index (SQI);

determine an amount of ghosting in the at least one dynamic region based on a dynamic HDR quality index (DQI);

weight at least one artifact in the at least one static region and the at least one dynamic region through a visual saliency map;

compute at least one edge preservation score based on the SQI and the DQI;

generate an output HDR quality metric by modulating the at least one computed edge preservation score on an input HDR quality metric; and

estimate the perceptual quality of the HDR image based on the generated output HDR quality metric.

9 . The electronic device as claimed in claim 8 , wherein the perceptual quality and estimation controller is configured to feedback the generated output HDR quality metric for tuning at least one parameter to optimize the perceptual quality of the HDR image.

10 . The electronic device as claimed in claim 8 , wherein the SQI determines an HDR blur through at least one lost LDR edge in the at least one static region of the HDR image due to one of an overexposure or an underexposure.

11 . The electronic device as claimed in claim 8 , wherein the DQI determines a HDR ghosting through additional edges in the at least one dynamic region of the HDR image due to a foreground motion.

12 . The electronic device as claimed in claim 8 , wherein the visual saliency map is used to weigh an edge-cost in the HDR image compared to at least one the LDR image before computing the edge preservation score, wherein the edge-cost corresponds to edge loss and a presence of extra edges in the HDR image.

13 . The electronic device as claimed in claim 8 , wherein the SQI is generated by:

accumulating all edges present in the at least one static region of an input LDR sequence into one edge map;

detecting a blur in an output HDR image by computing a number of edge pixels in the accumulated edges that are not found in the output HDR image at corresponding location;

computing the blur as a fraction of edge pixels in the accumulated edge map that are not present in the output HDR image; and

subtracting the computed blur ratio from a predefined value to obtain the SQI, wherein the predefined value is 1.

14 . The electronic device as claimed in claim 8 , wherein the DQI is generated by:

counting a number of edge pixels in the at least one dynamic region for each of LDR images in the input sequence;

determining the LDR image containing a maximum number of edges;

accumulating the edges from the LDR image and the edges in the output HDR image;

detecting the at least one artifact as extra edges present in the output HDR image that are not present in the input LDR image with the maximum number of edges; and

computing the DQI as the maximum number of edge pixels in the input sequence divided by the number of edges in the accumulated edge map, wherein the computation is always less than or equal to one, where the computation is equal to one if there are no extra edge pixels in the output HDR image.

15 . A perceptual quality and estimation control circuit configured to process a plurality of low dynamic range (LDR) images of different exposures and generate a high dynamic range (HDR) image, the control circuit comprising:

a dynamic region detection circuit configured to divide an image into at least one static region and at least one dynamic region;

a static HDR quality index (SQI) circuit configured to determine at least one preserved edge in the at least one static region based on an SQI;

a dynamic HDR quality index (DQI) circuit configured to determine an amount of ghosting in the at least one dynamic region based on a DQI;

a visual saliency prediction circuit configured to weight at least one artifact in the at least one static region and the at least one dynamic region through a visual saliency map;

a combinatorial circuit configured to compute at least one edge preservation score based on the SQI and the DQI;

a convolutional circuit configured to generate an output HDR quality metric by modulating the at least one computed edge preservation score on an input HDR quality metric; and

a feedback circuit configured to estimate and compensate the perceptual quality of the HDR image based on the generated output HDR quality metric.

16 . The perceptual quality and estimation control circuit as claimed in claim 15 , wherein the feedback circuit is configured to feedback the generated output HDR quality metric for tuning at least one parameter to optimize the perceptual quality of the HDR image.

17 . The perceptual quality and estimation control circuit as claimed in claim 15 , wherein the SQI circuit determines an HDR blur through at least one lost LDR edge in the at least one static region of the HDR image due to one of an overexposure or an underexposure.

18 . The perceptual quality and estimation control circuit as claimed in claim 15 ,

wherein the DQI circuit determines an HDR ghosting through additional edges in the at least one dynamic region of the HDR image due to a foreground motion,

wherein the visual saliency prediction circuit weighs an edge-cost in the HDR image compared to at least one of the plurality of LDR images before computing the edge preservation score, wherein the edge-cost corresponds to edge loss and a presence of extra edges in the HDR image.

19 . The perceptual quality and estimation control circuit as claimed in claim 15 , wherein the SQI circuit comprises:

an accumulator configured to accumulate all edges present in the at least one static region of an input LDR sequence into one edge map;

an edge detector configured to detect a blur in an output HDR image by computing a number of edge pixels in the accumulated edges that are not found in the output HDR image at corresponding location; and

a processor configured to compute the blur as a fraction of edge pixels in the accumulated edge map that are not present in the output HDR image, and subtract the computed blur ratio from a predefined value to obtain the SQI.

20 . The perceptual quality and estimation control circuit as claimed in claim 15 , wherein the DQI circuit comprises:

a counter configured to count a number of edge pixels in the at least one dynamic regions for each of LDR images in the input sequence;

a maximizer configured to determine the LDR image containing a maximum number of edges;

an accumulator configured to accumulate the edges from the LDR image and the edges in the output HDR image;

an edge detector configured to detect the at least one artifact as extra edges present in the output HDR image that are not present in the input LDR image with the maximum number of edges; and

a processor configured to compute the DQI as the maximum number of edge pixels in the plurality of LDR images divided by the number of edges in the accumulated edge map.

Assignments (1)

Priority Claims (1)

Continuity (1)

References Cited (47)