DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Specification
The title of the invention is not descriptive. A new title is required that is clearly indicative of the invention to which the claims are directed.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-3, 5, 8 and 11-13 are rejected under 35 U.S.C. 103 as being unpatentable over Yang et al. US-PGPUB No. 2023/0153510 (hereinafter Yang) in view of Park et al. US-PGPUB No. 2022/0327692 (hereinafter Park); Perazzi et al. US-PGPUB No. 2022/0122224 (hereinafter Perazzi) and Yadid-Pecht et al. US-PGPUB No. 2021/0217151 (hereinafter Pecht).
Re Claim 1:
Yang in view of Park teaches an image generation apparatus comprising at least one first processor, the at least one first processor being configured to:
acquire a first image (Yang teaches at FIG. 1A and Paragraph 0025 acquiring a mask image 106 by receiving a mask image 105);
enhance or extract mutually different frequency components in the first image to generate a plurality of frequency-processed images (Yang teaches at FIG. 1A extracting low-frequency component features 112 (feature map) and high-frequency component features (feature map) which are high-frequency image and low-frequency image according to Park at FIG. 11 and Paragraph 0150 describing that the high-frequency component features as high frequency image and low-frequency features as low-frequency image);
perform different computational processing on each of the plurality of frequency- processed images (Yang teaches at FIG. 1A performing different computation processing using low-frequency extraction branch 110 and high-frequency extraction branch 115);
synthesize respective frequency components of the plurality of frequency-processed images on which the computational processing is performed to generate at least one second image (Yang teaches at FIG. 1A and Paragraph 0061 using reconstruction unit 120 to synthesize the high-frequency component features and low-frequency component features to generate estimated fabrication image 125 as an output image).
Yang at least suggests the claim limitation:
output the first image and the at least one second image as training data used for machine learning of a mathematical model that performs predetermined inference on an input image (Yang teaches at Paragraph 0130 that images generated applying one or more of the techniques may be used to train, test or certify DNNs used to recognize objects and environments in the real world. Yang teaches at Paragraph 0137 that training data might include a set of images that each includes a representation of a type of object, where each image also includes a label).
Perazzi teaches the claim limitation:
output the first image and the at least one second image as training data used for machine learning of a mathematical model that performs predetermined inference on an input image (Perazzi teaches at FIG. 3 and Paragraph 0051-0053 outputting the digital image 302 and retouched digital image 332 as training data for training the neural network. Perazzi teaches at Paragraph 01076 that image retouching system 106 trains the skin tone correction neural network 318 utilizing training data 602 that includes image pairs of input images 604 and the corresponding retouched images 606).
Moreover, Perazzi teaches an image generation apparatus comprising at least one first processor, the at least one first processor being configured to:
acquire a first image (Perazzi teaches at FIG. 4 acquiring a digital image 302);
enhance or extract mutually different frequency components in the first image to generate a plurality of frequency-processed images (Perazzi teaches at FIG. 5 that high-frequency band is a high-frequency image, the mid-frequency band is the mid-frequency image and low-frequency band 312 is a low-frequency image. Perazzi teaches at Paragraph 0097 that mid-frequency band 314 is an image that shows areas that differ between the mid-frequency layer 504 and the modified mid-frequency layer 510);
perform different computational processing on each of the plurality of frequency- processed images (Perazzi teaches at FIG. 3 and Paragraph 0053 using skin tone correction neural network to obtain modified low-frequency image and using skin correction neural network to obtain modified high-frequency image);
synthesize respective frequency components of the plurality of frequency-processed images on which the computational processing is performed to generate at least one second image (Perazzi teach at FIG. 3 and Paragraph 0019 using combiner 330 to obtain retouched digital image 332); and
output the first image and the at least one second image as training data used for machine learning of a mathematical model that performs predetermined inference on an input image (Perazzi teaches at FIG. 3 and Paragraph 0051-0053 outputting the digital image 302 and retouched digital image 332 as training data for training the neural network. Perazzi teaches at Paragraph 01076 that image retouching system 106 trains the skin tone correction neural network 318 utilizing training data 602 that includes image pairs of input images 604 and the corresponding retouched images 606).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have incorporated the image pairs as the training dataset as taught in Perazzi FIG. 6 into Yang’s training dataset to have trained the neural networks of Yang using the image pairs as the training dataset. One of the ordinary skill in the art would have been motivated to have provided the training dataset to have inferenced/tuned the weights of the neural network.
Moreover, Pecht teaches an image generation apparatus comprising at least one first processor, the at least one first processor being configured to:
acquire a first image (Pecht teaches at FIG. 1 and Paragraph 0059 acquiring a wide dynamic range image 20);
enhance or extract mutually different frequency components in the first image to generate a plurality of frequency-processed images (Pecht teaches at FIG. 1 and Paragraph 0085 that this result can then be saved and then the input to the second level can be fed back to the first set of modules with different weights. Pecht teaches at FIGS. 1-3 separate neural network layers with different weights for outputting the high-frequency-processed image by the other layers of the neural networks with weights provided at Paragraph 0078 and low-frequency processed image by the first layer of FIG. 2 with the weights provided at Paragraph 0075);
perform different computational processing on each of the plurality of frequency- processed images (Pecht teaches at FIG. 1 and Paragraph 0085 that this result can then be saved and then the input to the second level can be fed back to the first set of modules with different weights. Pecht teaches at FIGS. 1-3 separate neural network layers with different weights for outputting the high-frequency-processed image by the other layers of the neural networks with weights provided at Paragraph 0078 and low-frequency processed image by the first layer of FIG. 2 with the weights provided at Paragraph 0075. Pecht teaches at Paragraph 0080 that the weights and parameters for each neuron in the neural network can be adjusted);
synthesize respective frequency components of the plurality of frequency-processed images on which the computational processing is performed to generate at least one second image (Perazzi teach at Paragraph 0081-0082 that the neural network 70 will generate the final result 80 from the coarse low dynamic range image 60 and the resulting system can produce LDR image from a corresponding WDR input image and at Paragraph 0089-0090 that the first transition image and the final transition image are combined to produce the final low dynamic range image); and
output the first image and the at least one second image as training data used for machine learning of a mathematical model that performs predetermined inference on an input image (Pecht teaches at claim 4 and Paragraph 0072 that said neural network is trained using a selected training set of input training wide dynamic range images and corresponding output images).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have incorporated the image pairs as the training dataset as taught in Pecht at claim 4 into Yang’s training dataset to have trained the neural networks of Yang using the image pairs as the training dataset. One of the ordinary skill in the art would have been motivated to have provided the training dataset to have inferenced/tuned the weights of the neural network.
Re Claim 2:
The claim 2 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that the at least one first processor is configured to perform, as the computational processing, processing for applying mutually different weight coefficients to the plurality of frequency-processed images.
However, Perazzi and Pecht teach the claim limitation that the at least one first processor is configured to perform, as the computational processing, processing for applying mutually different weight coefficients to the plurality of frequency-processed images (
Pecht teaches at FIG. 1 and Paragraph 0085 that this result can then be saved and then the input to the second level can be fed back to the first set of modules with different weights. Pecht teaches at FIGS. 1-3 separate neural network layers with different weights for outputting the high-frequency-processed image by the other layers of the neural networks with weights provided at Paragraph 0078 and low-frequency processed image by the first layer of FIG. 2 with the weights provided at Paragraph 0075. Pecht teaches at Paragraph 0080 that the weights and parameters for each neuron in the neural network can be adjusted.
Perazzi teaches at Paragraph 0112 that the skin tone correction loss model 612 compares the retouched low-frequency layer 610 with the low-frequency retouched image 614 to determine a reconstruction error loss and back propagates the reconstruction error loss to tune the skin tone correction neural network 318. In response, the skin tone correction neural network 318 updates weights and parameters to minimize the reconstruction error loss. Indeed, the image retouching system 106 iteratively tunes and trains the skin tone correction neural network 318 to learn a set of best-fit parameters that corrects skin tones within images.
Perazzi teaches at Paragraph 0117 that the image retouching system 106 pre-processes the training data 602 to generate the mid-frequency input band 616 and the mid-frequency retouched band 622. Further, the image retouching system 106 provides a skin mask corresponding to a training data image pair to the blemish smoothing neural network 322. Then, the blemish smoothing neural network 322 processes the mid-frequency input band 616 to generate the modified mid-frequency band 618, which is provided to a blemish smoothing loss model 620. The blemish smoothing loss model 620 determines an amount of reconstruction error loss (e.g., based on the mid-frequency retouched band 622), which is backpropagated to the blemish smoothing neural network 322 for updating of neural network parameters.
In the same manner, the weights of the skin correction neural network 326 are adjusted using the high-frequency input band).
Re Claim 3:
The claim 3 encompasses the same scope of inventio as that of the claim 2 except additional claim limitation that the at least one first processor is configured to:
generate a first frequency-processed image including relatively low frequency components and a second frequency-processed image including relatively high frequency components; and
perform, as the computational processing, processing for applying a relatively small weight coefficient to the first frequency-processed image and applying a relatively large weight coefficient to the second frequency-processed image.
However, Perazzi and Pecht teach the claim limitation that the at least one first processor is configured to:
generate a first frequency-processed image including relatively low frequency components and a second frequency-processed image including relatively high frequency components; and
perform, as the computational processing, processing for applying a relatively small weight coefficient to the first frequency-processed image and applying a relatively large weight coefficient to the second frequency-processed image (
Pecht teaches at FIG. 1 and Paragraph 0085 that this result can then be saved and then the input to the second level can be fed back to the first set of modules with different weights. Pecht teaches at FIGS. 1-3 separate neural network layers with different weights for outputting the high-frequency-processed image by the other layers of the neural networks with weights provided at Paragraph 0078 and low-frequency processed image by the first layer of FIG. 2 with the weights provided at Paragraph 0075. Pecht teaches at Paragraph 0080 that the weights and parameters for each neuron in the neural network can be adjusted.
Perazzi teaches at Paragraph 0112 that the skin tone correction loss model 612 compares the retouched low-frequency layer 610 with the low-frequency retouched image 614 to determine a reconstruction error loss and back propagates the reconstruction error loss to tune the skin tone correction neural network 318. In response, the skin tone correction neural network 318 updates weights and parameters to minimize the reconstruction error loss. Indeed, the image retouching system 106 iteratively tunes and trains the skin tone correction neural network 318 to learn a set of best-fit parameters that corrects skin tones within images.
Perazzi teaches at Paragraph 0117 that the image retouching system 106 pre-processes the training data 602 to generate the mid-frequency input band 616 and the mid-frequency retouched band 622. Further, the image retouching system 106 provides a skin mask corresponding to a training data image pair to the blemish smoothing neural network 322. Then, the blemish smoothing neural network 322 processes the mid-frequency input band 616 to generate the modified mid-frequency band 618, which is provided to a blemish smoothing loss model 620. The blemish smoothing loss model 620 determines an amount of reconstruction error loss (e.g., based on the mid-frequency retouched band 622), which is backpropagated to the blemish smoothing neural network 322 for updating of neural network parameters.
In the same manner, the weights of the skin correction neural network 326 are adjusted using the high-frequency input band).
Re Claim 5:
The claim 5 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that the at least one first processor is configured to:
perform filter processing on the first image to generate a first frequency-processed image; and
subtract frequency components of the first frequency-processed image from the first image to generate a second frequency-processed image.
However, Perazzi teaches the claim limitation that the at least one first processor is configured to:
perform filter processing on the first image to generate a first frequency-processed image; and
subtract frequency components of the first frequency-processed image from the first image to generate a second frequency-processed image (Perazzi teaches at Paragraph 0099 that the second upsampler 514 utilizes a bilinear kernel to generate the modified high-frequency layer 516 (modified mid-frequency layer). Perazzi teaches at FIG. 5 and Paragraph 0098 that the high-frequency band 316 is obtained using the second differentiator 518 by comparing the modified mid-frequency layer with the face portrait 308).
Re Claim 8:
The claim 8 recites a learning apparatus comprising at least one second processor, the at least one second processor being configured to train the mathematical model using, as training data, the first image and the at least one second image provided from the image generation apparatus according to claim 1.
Perazzi and Pecht further teach the claim limitation:
a learning apparatus comprising at least one second processor, the at least one second processor being configured to train the mathematical model using, as training data, the first image and the at least one second image provided from the image generation apparatus according to claim 1 (Perazzi teaches at FIG. 3 and Paragraph 0051-0053 outputting the digital image 302 and retouched digital image 332 as training data for training the neural network. Perazzi teaches at Paragraph 01076 that image retouching system 106 trains the skin tone correction neural network 318 utilizing training data 602 that includes image pairs of input images 604 and the corresponding retouched images 606.
Pecht teaches at claim 4 that said neural network is trained using a selected training set of input training wide dynamic range images and corresponding output images).
Re Claim 11:
The claim 11 encompasses the same scope of inventio as that of the claim 10 except additional claim limitation that the at least one third processor is configured to:
acquire the input image;
enhance or extract a specific frequency component in the input image to generate a frequency-processed image;
input the input image to the mathematical model to acquire a first inference result;
input the frequency-processed image to the mathematical model to acquire a second inference result; and
output the detection result by comprehensively evaluating the first inference result and the second inference result.
Moreover, Perazzi teaches the at least one third processor is configured to:
acquire the input image (Perazzi teaches at FIG. 4 acquiring a digital image 302);
enhance or extract a specific frequency component in the input image to generate a frequency-processed image (Perazzi teaches at FIG. 5 that high-frequency band is a high-frequency image, the mid-frequency band is the mid-frequency image and low-frequency band 312 is a low-frequency image. Perazzi teaches at Paragraph 0097 that mid-frequency band 314 is an image that shows areas that differ between the mid-frequency layer 504 and the modified mid-frequency layer 510);
input the input image to the mathematical model to acquire a first inference result;
input the frequency-processed image to the mathematical model to acquire a second inference result; and
(Perazzi teaches at FIG. 3 and Paragraph 0053 using skin tone correction neural network to obtain modified low-frequency image and using skin correction neural network to obtain modified high-frequency image);
output the detection result by comprehensively evaluating the first inference result and the second inference result (Perazzi teach at FIG. 3 and Paragraph 0019 using combiner 330 to obtain retouched digital image 332.
Perazzi teaches at FIG. 3 and Paragraph 0051-0053 outputting the digital image 302 and retouched digital image 332 as training data for training the neural network. Perazzi teaches at Paragraph 01076 that image retouching system 106 trains the skin tone correction neural network 318 utilizing training data 602 that includes image pairs of input images 604 and the corresponding retouched images 606).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have incorporated the image pairs as the training dataset as taught in Perazzi FIG. 6 into Yang’s training dataset to have trained the neural networks of Yang using the image pairs as the training dataset. One of the ordinary skill in the art would have been motivated to have provided the training dataset to have inferenced/tuned the weights of the neural network.
Moreover, Pecht teaches the claim limitation that the at least one third processor is configured to:
acquire the input image;
(Pecht teaches at FIG. 1 and Paragraph 0059 acquiring a wide dynamic range image 20);
enhance or extract a specific frequency component in the input image to generate a frequency-processed image (Pecht teaches at FIG. 1 and Paragraph 0085 that this result can then be saved and then the input to the second level can be fed back to the first set of modules with different weights. Pecht teaches at FIGS. 1-3 separate neural network layers with different weights for outputting the high-frequency-processed image by the other layers of the neural networks with weights provided at Paragraph 0078 and low-frequency processed image by the first layer of FIG. 2 with the weights provided at Paragraph 0075);
input the input image to the mathematical model to acquire a first inference result;
input the frequency-processed image to the mathematical model to acquire a second inference result; and (Pecht teaches at FIG. 1 and Paragraph 0085 that this result can then be saved and then the input to the second level can be fed back to the first set of modules with different weights. Pecht teaches at FIGS. 1-3 separate neural network layers with different weights for outputting the high-frequency-processed image by the other layers of the neural networks with weights provided at Paragraph 0078 and low-frequency processed image by the first layer of FIG. 2 with the weights provided at Paragraph 0075. Pecht teaches at Paragraph 0080 that the weights and parameters for each neuron in the neural network can be adjusted);
output the detection result by comprehensively evaluating the first inference result and the second inference result (Perazzi teach at Paragraph 0081-0082 that the neural network 70 will generate the final result 80 from the coarse low dynamic range image 60 and the resulting system can produce LDR image from a corresponding WDR input image and at Paragraph 0089-0090 that the first transition image and the final transition image are combined to produce the final low dynamic range image.
Pecht teaches at claim 4 and Paragraph 0072 that said neural network is trained using a selected training set of input training wide dynamic range images and corresponding output images).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have incorporated the image pairs as the training dataset as taught in Pecht at claim 4 into Yang’s training dataset to have trained the neural networks of Yang using the image pairs as the training dataset. One of the ordinary skill in the art would have been motivated to have provided the training dataset to have inferenced/tuned the weights of the neural network.
Re Claim 12:
The claim 12 recites an image generation method comprising processing performed by at least one first processor that an image generation apparatus has, the processing comprising: acquiring a first image;
enhancing or extracting mutually different frequency components in the first image to generate a plurality of frequency-processed images;
performing different computational processing on each of the plurality of frequency- processed images;
synthesizing respective frequency components of the plurality of frequency-processed images on which the computational processing is performed to generate at least one second image; and
outputting the first image and the at least one second image as training data used for machine learning of a mathematical model that performs predetermined inference on an input image.
The claim 12 is in parallel with the claim 1 in a method form. The claim 12 is subject to the same rationale of rejection as the claim 1.
Re Claim 13:
Claim 13 recites a learning method comprising processing performed by at least one second processor that a learning apparatus has, the processing comprising
training the mathematical model using, as training data, the first image and the at least one second image provided using the image generation method according to claim 12.
Perazzi and Pecht further teach the claim limitation:
a learning method comprising processing performed by at least one second processor that a learning apparatus has, the processing comprising
training the mathematical model using, as training data, the first image and the at least one second image provided using the image generation method according to claim 12 (Perazzi teaches at FIG. 3 and Paragraph 0051-0053 outputting the digital image 302 and retouched digital image 332 as training data for training the neural network. Perazzi teaches at Paragraph 01076 that image retouching system 106 trains the skin tone correction neural network 318 utilizing training data 602 that includes image pairs of input images 604 and the corresponding retouched images 606.
Pecht teaches at claim 4 and Paragraph 0072 that said neural network is trained using a selected training set of input training wide dynamic range images and corresponding output images).
Claim 4 is rejected under 35 U.S.C. 103 as being unpatentable over Yang et al. US-PGPUB No. 2023/0153510 (hereinafter Yang) in view of Park et al. US-PGPUB No. 2022/0327692 (hereinafter Park); Perazzi et al. US-PGPUB No. 2022/0122224 (hereinafter Perazzi); Yadid-Pecht et al. US-PGPUB No. 2021/0217151 (hereinafter Pecht); Takeda US-PGPUB No. 2022/0207723 (hereinafter Takeda); Yu et al. US-PGPUB No. 2024/0273672 (hereinafter Yu); Zhu et al. US-PGPUB No. 2022/0321852 (hereinafter Zhu).
Re Claim 4:
The claim 4 encompasses the same scope of inventio as that of the claim 1 except additional claim limitation that the at least one first processor is configured to perform, as the computational processing, processing on each of the plurality of frequency-processed images to enlarge a difference from an average value of brightness values for each pixel by a different magnification factor.
Zhu discloses at Paragraph 0121 “an intensity/brightness of each of the plurality of pixels in the dehazed image to conform to a desired range of brightness in the input image”. The intensity value of a pixel is proportional to a brightness value of the pixel.
Yu thus further teaches the claim limitation that the at least one first processor is configured to perform, as the computational processing, processing on each of the plurality of frequency-processed images to enlarge a difference from an average value of brightness values for each pixel by a different magnification factor (Yu teaches at Paragraph 0042-0045 that the intensity/brightness of the input/output images are normalized and scaled for each pixel of the image using mean square error loss function to have tuned the machine learning model).
Takeda US-PGPUB No. 2022/0207723 (hereinafter Takeda) teaches at FIG. 5 that the x-ray image 10 is decomposed into low-frequency component image and high-frequency component image wherein the low frequency component image 12 contains a lot of color information and structural information such as the shape of the object.
Takeda teaches the claim limitation the at least one first processor is configured to perform, as the computational processing, processing on each of the plurality of frequency-processed images to enlarge a difference from an average value of brightness values for each pixel by a different magnification factor (Takeda teaches at Paragraph 0096 he high-resolution image generation unit 2b generates the high-resolution high-frequency component image from the high-frequency component image 11 by the learning model 40.
Takeda teaches at Paragraph 0116 that the image enlargement unit 2d is configured to enlarge the low-frequency component image 12 acquired by performing the frequency resolution processing by the frequency resolution processing unit 2a to generate the low-frequency component enlarged image 14. The learning model 40 is further trained to enlarge the image to be generated when enhancing the resolution of the image. The high-resolution image generation unit 2b is configured to generate the high-resolution high-frequency component enlarged image 15 improved in resolution than the high-frequency component image 11 and enlarged in size from the high-frequency component image 11 by the learning model 40. The image synthesis unit 2c is configured to synthesize the low-frequency component enlarged image 14 and the high-resolution high-frequency component enlarged image 15 to generate the high-resolution X-ray image 13).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have processed a x-ray image of Takeda or the digital image of Yu as the input image and to have trained the neural network architecture of Yang, Perazzi and Pecht utilizing brightness/intensity scaling/amplification/normalization to have obtained an enhanced output image by the trained neural network models. One of the ordinary skill in the art would have been motivated to have trained a neural network utilizing the normalized input and output iamges.
Claims 6 and 9 are rejected under 35 U.S.C. 103 as being unpatentable over Yang et al. US-PGPUB No. 2023/0153510 (hereinafter Yang) in view of Park et al. US-PGPUB No. 2022/0327692 (hereinafter Park); Perazzi et al. US-PGPUB No. 2022/0122224 (hereinafter Perazzi); Yadid-Pecht et al. US-PGPUB No. 2021/0217151 (hereinafter Pecht); Yu et al. US-PGPUB No. 2024/0273672 (hereinafter Yu); Zhu et al. US-PGPUB No. 2022/0321852 (hereinafter Zhu);
Amthor et al. US-PGPUB No. 2021/0356729 (hereinafter Amthor); Chen US-PGPUB No. 2022/0130016 (hereinafter Chen).
Re Claim 6:
The claim 6 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that the at least one first processor is configured to perform normalization processing on the first image and the at least one second image to normalize brightness.
Zhu discloses at Paragraph 0121 “an intensity/brightness of each of the plurality of pixels in the dehazed image to conform to a desired range of brightness in the input image”. The intensity value of a pixel is proportional to a brightness value of the pixel.
Yu thus further teaches the claim limitation that the at least one first processor is configured to perform normalization processing on the first image and the at least one second image to normalize brightness (Yu teaches at Paragraph 0042-0045 that the intensity/brightness of the input/output images are normalized).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have normalized the input and output image as the image pair can be more balanced. One of the ordinary skill in the art would have been motivated to have normalized the input and output images to have trained the neural networks based on the normalized input and output images.
Amthor et al. US-PGPUB No. 2021/0356729 (hereinafter Amthor)/Chen US-PGPUB No. 2022/0130016 (hereinafter Chen) further teaches the claim limitation that the at least one first processor is configured to perform normalization processing on the first image and the at least one second image to normalize brightness (Amthor teaches at Paragraph 0044 performing a brightness normalization operation on the input image and Chen teaches at Paragraph 0052 the performing a brightness normalization operation steps on a new synthesized fingerprint image).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have normalized the input and output image as the image pair can be more balanced. One of the ordinary skill in the art would have been motivated to have normalized the input and output images to have trained the neural networks based on the normalized input and output images.
Re Claim 9:
The claim 9 encompasses the same scope of inventio as that of the claim 8 except additional claim limitation that the at least one second processor is configured to perform normalization processing on the first image and the at least one second image to normalize brightness.
Zhu discloses at Paragraph 0121 “an intensity/brightness of each of the plurality of pixels in the dehazed image to conform to a desired range of brightness in the input image”. The intensity value of a pixel is proportional to a brightness value of the pixel.
Yu thus further teaches the claim limitation that the at least one first processor is configured to perform normalization processing on the first image and the at least one second image to normalize brightness (Yu teaches at Paragraph 0042-0045 that the intensity/brightness of the input/output images are normalized).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have normalized the input and output image as the image pair can be more balanced. One of the ordinary skill in the art would have been motivated to have normalized the input and output images to have trained the neural networks based on the normalized input and output images.
Amthor et al. US-PGPUB No. 2021/0356729 (hereinafter Amthor)/Chen US-PGPUB No. 2022/0130016 (hereinafter Chen) further teaches the claim limitation that the at least one second processor is configured to perform normalization processing on the first image and the at least one second image to normalize brightness (Amthor teaches at Paragraph 0044 performing a brightness normalization operation on the input image and Chen teaches at Paragraph 0052 the performing a brightness normalization operation steps on a new synthesized fingerprint image).
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have normalized the input and output image as the image pair can be more balanced. One of the ordinary skill in the art would have been motivated to have normalized the input and output images to have trained the neural networks based on the normalized input and output images.
Claims 7, 10 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Yang et al. US-PGPUB No. 2023/0153510 (hereinafter Yang) in view of Park et al. US-PGPUB No. 2022/0327692 (hereinafter Park); Perazzi et al. US-PGPUB No. 2022/0122224 (hereinafter Perazzi); Yadid-Pecht et al. US-PGPUB No. 2021/0217151 (hereinafter Pecht); Wang et al. US-PGPUB No. 2023/0025557 (hereinafter Wang); Takeda US-PGPUB No. 2022/0207723 (hereinafter Takeda) and Albiol Colomer et al. US-PGPUB No.2018/0144501 (hereinafter Colomer).
Re Claim 7:
The claim 7 encompasses the same scope of invention as that of the claim 1 except additional claim limitation that the first image is a radiographic image including an image of a specific structural part of a target object in a specific frequency domain, and the mathematical model is a model that detects the image of the specific structural part included in the radiographic image.
Colomer discloses in Abstract that the x-ray image is a radiographic image.
Wang et al. US-PGPUB No. 2023/0025557 (hereinafter Wang) teaches at Paragraph 0059 that the low frequency components of the low dynamic range image, and the low frequency components contain a lot of color information and structural information (such as the shape of the object.
Takeda US-PGPUB No. 2022/0207723 (hereinafter Takeda) teaches at FIG. 5 that the x-ray image 10 (a radiographic image) is decomposed into low-frequency component image and high-frequency component image wherein the low frequency component image 12 contains a lot of color information and structural information such as the shape of the object.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have processed a x-ray image of Takeda as the input image by the neural network architecture of Yang, Perazzi and Pecht to have processed a radiographic image to obtain an enhanced radiographic image. One of the ordinary skill in the art would have been motivated to have processed a radiographic image.
Re Claim 10:
The claim 10 recites an image processing apparatus comprising at least one third processor, the at least one third processor being configured to output a detection result of an image of a specific structural part of a target object for an input image using the mathematical model trained by the learning apparatus according to claim 8.
Colomer discloses in Abstract that the x-ray image is a radiographic image.
Wang et al. US-PGPUB No. 2023/0025557 (hereinafter Wang) teaches at Paragraph 0059 that the low frequency components of the low dynamic range image, and the low frequency components contain a lot of color information and structural information (such as the shape of the object.
Takeda US-PGPUB No. 2022/0207723 (hereinafter Takeda) teaches at FIG. 5 that the x-ray image 10 (a radiographic image) is decomposed into low-frequency component image and high-frequency component image wherein the low frequency component image 12 contains a lot of color information and structural information such as the shape of the object.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have processed a x-ray image of Takeda as the input image by the neural network architecture of Yang, Perazzi and Pecht to have processed a radiographic image to obtain an enhanced radiographic image. One of the ordinary skill in the art would have been motivated to have processed a radiographic image.
Re Claim 14:
The claim 14 recites an image processing method comprising processing performed by at least one third processor that an image processing apparatus has, the processing comprising outputting a detection result of an image of a specific structural part of a target object for an input image using the mathematical model trained by the learning method according to claim 13.
Wang et al. US-PGPUB No. 2023/0025557 (hereinafter Wang) teaches at Paragraph 0059 that the low frequency components of the low dynamic range image, and the low frequency components contain a lot of color information and structural information (such as the shape of the object.
Takeda US-PGPUB No. 2022/0207723 (hereinafter Takeda) teaches at FIG. 5 that the x-ray image 10 is decomposed into low-frequency component image and high-frequency component image wherein the low frequency component image 12 contains a lot of color information and structural information such as the shape of the object.
It would have been obvious to one of the ordinary skill in the art before the filing date of the instant application to have processed a x-ray image of Takeda as the input image by the neural network architecture of Yang, Perazzi and Pecht to have processed a radiographic image to obtain an enhanced radiographic image. One of the ordinary skill in the art would have been motivated to have processed a radiographic image.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to JIN CHENG WANG whose telephone number is (571)272-7665. The examiner can normally be reached Mon-Fri 8:00-5:00.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, King Poon can be reached at 571-270-0728. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/JIN CHENG WANG/Primary Examiner, Art Unit 2617
Prosecution Insights