Method and apparatus for gradation correction and image edge extraction

• US 5,754,709 A
• Filed: 07/27/1995
• Issued: 05/19/1998
• Est. Priority Date: 11/10/1994
• Status: Expired due to Fees

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Imagawa, Taro, Koda, Toshiyuki, Yamamoto, Hiroshi, Maruno, Susumu, Moriya, Michiyo
• Primary Examiner(s): SHALWALA, BIPIN H
• Application Number:
  US08/508,037
• Time in Patent Office: 1,027 Days
• Field of Search: 382/190, 382/191, 382/192, 382/195, 382/199, 382/206, 382/227, 382/226, 382/274, 382/288, 382/156, 382/174, 382/173, 382/270, 382/272, 382/167, 358/521, 358/458, 358/455, 358/460-461, 358/462, 358/532, 364/276.6, 364/972.4, 395/21, 395/126, 348/251
• US Class Current: 382/274
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/20021 Dividing image into blocks, subimages or windows
• CPC Class Codes: G06T 7/12 Edge-based segmentation

Abstract

The present invention relates to an image processing apparatus in which gradation correction on each partial region is performed between an image input means and an edge extracting means so that an edge of a region within an image is accurately extracted. An image dividing means 2, which had already learned a relationship between the condition of an image and the necessity of image division so as to judge the condition of the image based on learning results and divide the image appropriately, divides an input image from an image input means 1 into a plurality of image blocks. An image correction information extracting means 3 calculates correction information on each image block. A gradation correction means 4, which had already learned a relationship between correction information and non-linear correction curves to judge correction information based on learning results and perform gradation correction on a target image block using a selected curve, performs gradation correction and enhances an edge which is to be extracted. An image synthesizing means 5 recombines the respective corrected image blocks. An edge extracting means 6 extracts an edge from a synthesized image. An image output means 7 outputs an edge image.

Claims

1. An image processing apparatus, comprising:

image input means for receiving an image frame/field;

image dividing means for dividing said image frame/field into a certain number of image blocks;

image correction information extracting means for extracting at least one type of image information for comparison with a predetermined reference indicating the presense of a section of an edge image signal within an image block;

gradation correction means for pretreating for edge extraction in accordance with the comparison between the image information and the predetermined reference by selecting a gradation correction for each of said image blocks;

image synthesizing means for synthesizing a pretreated frame/field by using the image signals each of which is pretreated by said gradation correction means;

edge extracting means for extracting an edge image from said pretreated frame/field; and

image output means for outputting said edge image.

3. An image processing apparatus in accordance with claim 1, wherein the image dividing means comprises:

an image block extracting part for extracting image blocks each of which has a region of m×

n pixels and image blocks each of which has a region of k×

l pixels (k≦

m, l≦

n) with a coordinate point (x,y) within said image frame/field as a center of gravity and for supplying the image blocks each of which has a region of m×

n pixels to said image correction information extracting means while supplying the image blocks each of which has a region of k×

l pixels to said gradation correction means;

a center of gravity setting part for calculating the center of gravity for the image blocks to be extracted next by said image block extracting part;

the center of gravity of said image block extracting part being renewed by the center of gravity which is calculated by said center of gravity setting part; and

a process of said image block extracting part being repeatedly performed.

5. An image processing apparatus in accordance with claim 1, wherein

said certain number of image blocks is one image block or a plurality of image blocks, and wherein

said image dividing means comprises:

an image division information extracting part for extracting at least one type of division information which serves as a reference of image division from a target image,

an image division judging part for determining whether to divide said target image in consideration of said division information, and

an image dividing part for dividing the target image into image blocks in accordance with a judgment yielded by said image division judging part, said target image being said image frame/field or the image blocks which are divided by said image dividing part.

7. An image processing apparatus in accordance with claim 5, wherein said image dividing means repeats a process of said image division information extracting part, said image division judging part and said image dividing part a certain times.

9. An image processing apparatus in accordance with claim 5, wherein said image division judging part comprises:

a division reference value memory part for storing division reference values corresponding to said division information; and

a comparison part for comparing said target image with each of said division reference values and for outputting "1" in case of division needed or "0" in case of division not needed to said image dividing part.

11. An image processing apparatus in accordance with claim 5, wherein said image division judging part comprises:

a learning type image division judging part for establishing a relationship between said division information and a necessity of division through learning and for determining whether to divide said target image on basis of said division information and a result which is obtained by establishing said relationship; and

a learning data memory part for storing combinations of said division information (characteristic data signals) and said necessity of division ("1" or "0") as data which is used for said learning.

13. An image processing apparatus in accordance with claim 11, wherein component recognition units are combined in a multi-layer hierarchic structure to form a network, and a threshold value processing unit is disposed in the bottom layer, said threshold value processing unit is formed by at least a plurality of path input terminals, an adder for adding up and input signals from said path input terminals, a threshold value processor for processing an output signal from said adder by threshold value processing and a path output terminal for outputting an output of said threshold value processor,

the learning type image division judging part is formed by disposing a teaching data input terminal at a signal input part of said component recognition unit located in a layer precedent to the bottom layer of said network so as to input teaching data which is stored in said learning data memory part and by disposing a characteristic data signal input terminal at a signal input part of said component recognition unit located in a layer other than the bottom layer and the layer precedent to the bottom layer so as to input a characteristic data signal which is stored in said learning data memory part, and

said component recognition unit is formed by:

a component unit formed by a signal input part, a quantizer for quantizing in accordance with an input signal supplied from said signal input part, one or a plurality of said path input terminals, a plurality of said path output terminals and a path selecting part which selectively connects said path input terminals and said path output terminals to select a path in accordance with outputs from and said quantizer;

a structure memory part for storing the quantization range of said quantizer and the number of quantization as quantization constants;

an inner state memory part for storing the average and the dispersion of input signals supplied to said component units and the number of inputting of the input signals as an inner state; and

a reproduction creating part for halving said quantization range based on said inner state stored and reproducing the component unit as it was before division.

15. An image processing apparatus in accordance with claim 1, wherein the image correction information extracting means comprises:

a luminance extracting part for extracting a luminance signal from the image signal corresponding to each of said image blocks;

a certain number of luminance level comparators for comparing said luminance signal with a level value having a certain range so as to output "1" or "0"; and

a certain number of counters for counting outputs from said luminance level comparators.

17. An image processing apparatus in accordance with claim 1, wherein the gradation correction means comprises:

a learning type non-linear correction curve selecting part for establishing through learning a relationship between said image information and a non-linear correction curve or a non-linear gamma curve which is to be used for gradation correction, judging said image information on basis of said learning and selecting said non-linear correction curve or said non-linear gamma curve; and

non-linear correction part for correcting the image signal corresponding to each of said images blocks by using the non-linear correction curve or the non-linear gamma curve selected by said learning type non-linear correction curve selecting part.

19. An image processing apparatus in accordance with claim 17, wherein component recognition units are combined in a multi-layer hierarchic structure to form a network, and a threshold value processing unit is disposed in the bottom layer, said threshold value processing unit is formed by at least a plurality of path input terminals, an adder for adding up input signals from said path input terminals, a threshold value processor for processing an output signal from said adder by threshold value processing and a threshold value processing unit for outputting an output of said threshold value processor, and

said component recognition unit is formed by:

a component unit formed by the learning type non-linear correction curve selecting part which is formed by disposing a teaching signal input terminal at a signal input part of said component recognition unit located in a layer precedent to said bottom layer of said network so as to input a teaching signal, said signal input part, the quantizer for quantizing in accordance with an input signal from said signal input part, one or a plurality of said path input terminals, one or a plurality of said path output terminals and a path selector for selectively connecting said path input terminals and said path output terminals to select a path in accordance with an output from said quantizer;

a structure memory part for storing the quantization range of said quantizer and the number of quantization as quantization constants;

an inner state memory part for storing the average and the dispersion of input signals supplied to said component units and the number of inputting of the input signals as an inner state; and

a reproduction creating part for halving said quantization range based on said inner state stored and reproducing the component unit as it was before division.

2. An image processing apparatus, comprising:

image input means for receiving an image frame/field;

image dividing means for dividing said image frame/field into a certain number of image blocks;

image correction information extracting means for extracting at least one type of image information for comparison with a predetermined reference indicating the presense of a section of an edge image signal within an image block;

gradation correction means for pretreating for edge extraction in accordance with the comparison between the image information and the predetermined reference by selecting a gradation correction for each of said image blocks;

edge extracting means for extracting an edge image from the image signal which is pretreated by said gradation correction means;

image synthesizing means for synthesizing an edge frame/field by using the edge images each of which is extracted by said edge extracting means; and

image output means for outputting said edge frame/field.

4. An image processing apparatus in accordance with claim 2, wherein the image dividing means comprises:

an image block extracting part for extracting image blocks each of which has a region of m×

n pixels and image blocks each of which has a region of k×

l pixels (k≦

m, l≦

n) with a coordinate point (x,y) within said image frame/field as a center of gravity and for supplying the image blocks each of which has a region of m×

n pixels to said image correction information extracting means while supplying the image blocks each of which has a region of k×

l pixels to said gradation correction means;

a center of gravity setting part for calculating the center of gravity for the image blocks to be extracted next by said image block extracting part;

the center of gravity of said image block extracting part being renewed by the center of gravity which is calculated by said center of gravity setting part; and

a process of said image block extracting part being repeatedly performed.

6. An image processing apparatus in accordance with claim 2, wherein

said certain number of image blocks is one image block or a plurality of image blocks, and wherein

said image dividing means comprises:

an image division information extracting part for extracting at least one type of division information which serves as a reference of image division from a target image,

an image division judging part for determining whether to divide said target image in consideration of said division information, and

an image dividing part for dividing the target image into image blocks in accordance with a judgment yielded by said image division judging part, said target image being said image frame/field or the image blocks which are divided by said image dividing part.

8. An image processing apparatus in accordance with claim 6, wherein said image dividing means repeats a process of said image division information extracting part, said image division judging part and said image dividing part a certain times.

10. An image processing apparatus in accordance with claim 6, wherein said image division judging part comprises:

a division reference value memory part for storing division reference values corresponding to said division information; and

a comparison part for comparing said target image with each of said division reference values and for outputting "1" in case of division needed or "0" in case of division not needed to said image dividing part.

12. An image processing apparatus in accordance with claim 6, wherein said image division judging part comprises:

a learning type image division judging part for establishing a relationship between said division information and a necessity of division through learning and for determining whether to divide said target image on basis of said division information and a result which is obtained by establishing said relationship; and

a learning data memory part for storing combinations of said division information (characteristic data signals) and said necessity of division ("1" or "0") as data which is used for said learning.

14. An image processing apparatus in accordance with claim 12, wherein component recognition units are combined in a multi-layer hierarchic structure to form a network, and a threshold value processing unit is disposed in the bottom layer, said threshold value processing unit is formed by at least a plurality of path input terminals, an adder for adding up and input signals from said path input terminals, a threshold value processor for processing an output signal from said adder by threshold value processing and a path output terminal for outputting an output of said threshold value processor,

the learning type image division judging part is formed by disposing a teaching data input terminal at a signal input part of said component recognition unit located in a layer precedent to the bottom layer of said network so as to input teaching data which is stored in said learning data memory part and by disposing a characteristic data signal input terminal at a signal input part of said component recognition unit located in a layer other than the bottom layer and the layer precedent to the bottom layer so as to input a characteristic data signal which is stored in said learning data memory part, and

said component recognition unit is formed by:

a component unit formed by a signal input part, a quantizer for quantizing in accordance with an input signal supplied from said signal input part, one or a plurality of said path input terminals, a plurality of said path output terminals and a path selecting part which selectively connects said path input terminals and said path output terminals to select a path in accordance with outputs from and said quantizer;

a structure memory part for storing the quantization range of said quantizer and the number of quantization as quantization constants;

an inner state memory part for storing the average and the dispersion of input signals supplied to said component units and the number of inputting of the input signals as an inner state; and

a reproduction creating part for halving said quantization range based on said inner state stored and reproducing the component unit as it was before division.

16. An image processing apparatus in accordance with claim 2, wherein the image correction information extracting means comprises:

a luminance extracting part for extracting a luminance signal from the image signal corresponding to each of said image blocks;

a certain number of luminance level comparators for comparing said luminance signal with a level value having a certain range so as to output "1" or "0"; and

a certain number of counters for counting outputs from said luminance level comparators.

18. An image processing apparatus in accordance with claim 2, wherein the gradation correction means comprises:

a learning type non-linear correction curve selecting part for establishing through learning a relationship between said image information and a non-linear correction curve or a non-linear gamma curve which is to be used for gradation correction, judging said image information on basis of said learning and selecting said non-linear correction curve or said non-linear gamma curve; and

non-linear correction part for correcting the image signal corresponding to each of said images blocks by using the non-linear correction curve or the non-linear gamma curve selected by said learning type non-linear correction curve selecting part.

20. An image processing apparatus in accordance with claim 18, wherein component recognition units are combined in a multi-layer hierarchic structure to form a network, and a threshold value processing unit is disposed in the bottom layer, said threshold value processing unit is formed by at least a plurality of path input terminals, an adder for adding up input signals from said path input terminals, a threshold value processor for processing an output signal from said adder by threshold value processing and a threshold value processing unit for outputting an output of said threshold value processor, and

said component recognition unit is formed by:

a component unit formed by the leaning type non-linear correction curve selecting part which is formed by disposing a teaching signal input terminal at a signal input part of said component recognition unit located in a layer precedent to said bottom layer of said network so as to input a teaching signal, said signal input part, the quantizer for quantizing in accordance with an input signal from said signal input part, one or a plurality of said path input terminals, one or a plurality of said path output terminals and a path selector for selectively connecting said path input terminals and said path output terminals to select a path in accordance with an output from said quantizer;

a structure memory part for storing the quantization range of said quantizer and the number of quantization as quantization constants;

an inner state memory part for storing the average and the dispersion of input signals supplied to said component units and the number of inputting of the input signals as an inner state; and

a reproduction creating part for halving said quantization range based on said inner state stored and reproducing the component unit as it was before division.

21. An image processing method for producing an edge image, said edge image defining edges within one of an image frame and an image field, comprising the steps of:

a) receiving one of said image frame and said image field;

b) dividing said one of said image frame and said image field into a selected number of image blocks, each of said image blocks having a corresponding image signal and a corresponding gradation;

c) extracting at least one type of image information from said image signal for comparison with a predetermined reference indicating a presense of a section of an edge image within an image block;

d) correcting gradation by

1) selecting a correction of gradation for each image signal in accordance with the comparison between the predetermined reference and said image information, and

2) pretreating each image signal for edge extraction responsive to the selected gradation correction;

e) synthesizing a pretreated image by using the pretreated image signals; and

f) extracting the edge image from said pretreated image.

23. An image processing method in accordance with claim 21 or claim 22, wherein the dividing step (b) further comprises the steps of:

1) targeting one of an image frame and an image field to define a targeted image;

2) determining, based on information in the targeted image, whether the targeted image should be divided, and

i) if no, then the targeted image forms one image block,

ii) if yes, then divide the targeted image into at least two divided images, and successively target each one of the divided images as the defined targeted image; and

3) repeating step (3) until all of the image blocks have been formed into the selected number of image blocks.

24. An image processing method in accordance with claim 21 or claim 22, wherein the dividing step (b) further comprises the steps of:

1) selecting each of the image blocks by defining a respective center of gravity;

2) defining a region for each selected image block, the region being one of a first type and of a second type;

3) processing by step (c) each image block having a region of the first type; and

4) processing by step (d) each image block having a region of the second type.

25. An image processing method in accordance with claim 21 or claim 22, wherein the extracting step (c) further comprises the steps of:

1) extracting a luminance signal from the image signal corresponding to each one of said image blocks; and

2) determining a luminance level value of each luminance signal, wherein the luminance level value is one type of image information.

26. An image processing method in accordance with claim 21 or claim 22, wherein the step of correcting gradation (d)(1) further comprises the steps of:

1) providing a plurality of non-linear correction curves;

2) establishing a relationship between each one of the plurality of correction curves and the image information;

3) selecting one of the plurality of correction curves by evaluating each one of the plurality of correction curves based on the relationship; and

4) correcting the image signal in the corresponding image block according to the information in the selected correction curve.

22. An image processing method for producing an edge image, said edge image defining edges within one of an image frame and an image field, comprising the steps of:

a) receiving one of an image frame and an image field;

b) dividing said one of said image frame and said image field into a selected number of image blocks, each of said image blocks having a corresponding image signal and a corresponding gradation;

c) extracting at least one type of image information from said image signal for comparison with a predetermined reference indicating a presense of a section of an edge image within an image block;

d) correcting gradation by

1) selecting a correction of gradation for each image signal in accordance with the comparison between the predetermined reference and said image information, and

2) pretreating each image signal for edge extraction responsive to the selected gradation correction;

e) extracting an edge signal from each of said pretreated image signals; and

f) synthesizing the edge image from each of said edge signals extracted by said edge extracting means.

27. An image processing system for producing an edge image, said edge image defining edges within one of an image frame and an image field, comprising:

means for receiving one of said image frame and said image field;

dividing means for dividing said one of said image frame and said image field into a selected number of image blocks, each of said image blocks having a corresponding image signal and a corresponding gradation;

image information extraction means for extracting at least one type of image information from said image signal for comparison with a predetermined reference indicating a presense of a section of an edge image within an image block;

gradation correction means;

1) for selecting a correction of gradation for each image signal in accordance with the comparison between the predetermined reference and said image information, and

2) for pretreating each image signal for edge extraction responsive to the selected gradation correction;

image synthesizing means for synthesizing a pretreated image by using the pretreated image signals; and

edge image extracting means for extracting the edge image from said pretreated image.

28. An image processing system for producing an edge image, said edge image defining edges within one of an image frame and an image field, comprising:

means for receiving one of an image frame and an image field;

dividing means for dividing said one of said image frame and said image field into a selected number of image blocks, each of said image blocks having a corresponding image signal and a corresponding gradation;

image information extracting means for extracting at least one type of image information from said image signal for comparison with a predetermined reference indicating a presense of a section of an edge image within an image block;

gradation correction means

1) for selecting a correction of gradation for each image signal in accordance with the comparison between the predetermined reference and said image information, and

2) for pretreating each image signal for edge extraction responsive to the selected gradation correction;

edge signal extracting means for extracting an edge signal from each of said pretreated image signals; and

edge image synthesizing means for synthesizing the edge image from each of said edge signals extracted by said edge extracting means.