Methods and systems for adjusting digital image characteristics

• US 20030185457A1
• Filed: 03/29/2002
• Published: 10/02/2003
• Est. Priority Date: 03/29/2002
• Status: Active Grant

• Current Assignee: Sharp Kabushiki Kaisha (Hon Hai Precision Industry Co., Ltd.)
• Original Assignee: Sharp Laboratories of America Incorporated (Hon Hai Precision Industry Co., Ltd.)
• Inventors: Campbell, Richard John
• Granted Patent: US 7,221,807 B2
• Time in Patent Office: Days
• US Class Current: 382/254
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 5/40 using histogram techniques
• CPC Class Codes: G06T 5/60 using machine learning, e.g. neural networks
• CPC Class Codes: G06T 5/92 based on global image properties
• CPC Class Codes: H04N 1/4074 using histograms
• CPC Class Codes: H04N 1/6005 with luminance or chrominance signals, e.g. LC1C2, HSL or YUV

Abstract

Embodiments of the present invention comprise methods and systems for automatically adjusting images to conform to preference data.

Claims

1. A method for adjusting image characteristics, said method comprising:

linking image characteristic data with corresponding image adjustment data wherein said image adjustment data is related to a preference of a human observer;

training a neural net to generate automatic image enhancement data using said image characteristic data linked to said image adjustment data; and

enhancing characteristics of an image using said image enhancement data, wherein said enhancing makes said image more closely comply with said human preference.

2. The method of claim 1 wherein said image characteristic data and said image adjustment data relate to image tone.

3. The method of claim 1 wherein said image characteristic data comprises an image characteristic histogram.

4. The method of claim 1 wherein said image characteristic data comprises a luminance histogram.

5. The method of claim 1 wherein said image characteristic data and said image adjustment data are linked through a database.

6. The method of claim 1 wherein said neural net comprises two neuron layers.

7. The method of claim 1 wherein said neural net comprises a tangent sigmoid transfer function in the first neuron layer.

8. A method for automatically reproducing subjective image characteristic preference choices of a human observer, said method comprising:

said observer indicating a preference between an original image and an adjusted image wherein said adjusted image differs from said original image by a change in at least one image characteristic;

recording said original image characteristics and said adjusted image differences;

repeating said observer indicating step and said recording step until a substantial data set is obtained; and

training a neural net to generate enhancement data for reproducing said observer preference indications.

9. The method of claim 8 wherein said observer indicating a preference comprises said observer utilizing an interactive system comprising display of said original image and said adjusted image and said observer can make interactive adjustments to said adjusted image to indicate a preferred adjustment to said original image.

10. The method of claim 9 wherein said interactive system further comprises a display of a characteristic histogram of said original image and a characteristic histogram of said adjusted image.

11. The method of claim 9 wherein said interactive adjustments are made by mapping image characteristics through a conversion curve.

12. A method for automatically adjusting image characteristics according to subjective human preferences, said method comprising:

linking image characteristic data with corresponding image adjustment data wherein said image adjustment data comprises a preferred adjustment to said image characteristic data;

training a neural net to generate automatic image enhancement data using said image characteristic data linked to said image adjustment data;

converting a subject image to a luminance-segregated format comprising a luminance channel;

creating a channel data histogram;

using said channel data histogram as input to said neural net;

forming a Look-Up Table (LUT) from said neural net image enhancement data, said LUT representing an enhancement curve for processing said subject image; and

applying said enhancement curve LUT to said subject image thereby creating a corrected image, which conforms more closely to said subjective human preferences.

13. A method for automatically adjusting image tone according to subjective human preferences, said method comprising:

linking an image luminance histogram with corresponding luminance adjustment data wherein said adjustment data comprises a preferred adjustment to said image luminance histogram;

training a neural net to generate automatic image enhancement data using said image luminance histogram data linked to said adjustment data;

converting a subject image to a luminance-segregated format comprising a luminance channel;

creating a subject image luminance channel histogram;

using said subject image luminance channel histogram as input to said neural net;

forming a Look-Up Table (LUT) from said neural net automatic image enhancement data, said LUT representing an enhancement curve for processing said subject image luminance channel histogram;

applying said enhancement curve LUT to said subject image luminance channel histogram thereby adjusting said subject image histogram to form an adjusted image luminance histogram; and

creating an adjusted image with said adjusted image luminance histogram, said adjusted image conforming more closely to said subjective human preferences than said subject image.

14. A system for automatically reproducing subjective image characteristic preference choices of a human observer, said system comprising:

an observer data collector for indicating an observer preference between an original image and an adjusted image wherein said adjusted image differs from said original image by a change in at least one image characteristic;

a recorder for recording said original image characteristics and said adjusted image differences; and

a neural net trained to output adjustment data that will reproduce said observer'"'"'s preferences, said neural net being trained with said image characteristics and said image differences;

an image enhancer for enhancing a subject image with said adjustment data from said neural net thereby creating a preferred image that conforms more closely to said observer preferences than said subject image.

15. A computer readable medium comprising instructions for automatically adjusting image characteristics according to subjective human preferences, said instructions comprising the acts of:

linking image characteristic data with corresponding image adjustment data wherein said image adjustment data comprises a preferred adjustment to said image characteristic data;

training a neural net to generate automatic image enhancement data using said image characteristic data linked to said image adjustment data;

converting a subject image to a luminance-segregated format comprising a luminance channel and at least one chrominance channel;

creating a channel data histogram;

using said channel data histogram as input to said neural net;

forming a Look-Up Table (LUT) from said neural net automatic image enhancement data, said LUT representing an enhancement curve for processing said subject image; and

applying said enhancement curve LUT to said subject image thereby creating an enhanced image, which conforms more closely to said subjective human preferences.

16. A computer data signal embodied in an electronic transmission, said signal having the function of automatically adjusting image characteristics according to subjective human preferences, said signal comprising instructions for:

linking image characteristic data with corresponding image adjustment data wherein said image adjustment data comprises a preferred adjustment to said image characteristic data;

training a neural net to generate automatic image enhancement data using said image characteristic data linked to said image adjustment data;

converting a subject image to a luminance-segregated format comprising a luminance channel and at least one chrominance channel;

creating a channel data histogram;

using said channel data histogram as input to said neural net;

forming a Look-Up Table (LUT) from said neural net automatic image enhancement data, said LUT representing an enhancement curve for processing said subject image; and

applying said enhancement curve LUT to said subject image thereby creating an enhanced image, which conforms more closely to said subjective human preferences.