Techniques for Reducing Distractions in an Image

Abstract

Techniques for reducing a distractor object in a first image are presented herein. A system can access a mask and the first image. A distractor object in the first image can be inside a region of interest and can have a pixel with an original attribute. Additionally, the system can process, using a machine-learned inpainting model, the first image and the mask to generate an inpainted image. The pixel of the distractor object in the inpainted image can have an inpainted attribute in chromaticity channels. Moreover, the system can determine a palette transform based on a comparison of the first image and the inpainted image. The transform attribute can be different from the inpainted attribute. Furthermore, the system can process the first image to generate a recolorized image. The pixel in the recolorized image can have a recolorized attribute based on the transform attribute of the palette transform.

1 .- 20 . (canceled) 
     
     
         21 . A computer-implemented method for reducing a distractor object in an image, the method comprising:
 accessing the image and a mask, the image being in a red-green-blue (RGB) color space, wherein the mask indicates a region of interest associated with the image, and wherein a distractor object is inside the region of interest and has a first pixel with a first RGB value;   processing, using a machine-learned inpainting model, the image and the mask to generate an inpainted image, wherein the first pixel of the distractor object has a first inpainted attribute value in one or more chromaticity channels;   modifying, using a voting technique, the first pixel of the distractor object to a second inpainted attribute value in the one or more chromaticity channels, the second inpainted attribute value being different than the first inpainted attribute value; and   processing the image to generate a final image in the RGB color space, wherein the first pixel of the distractor object has a second RGB value that is different than the first RGB value, the second RGB value being based on the second inpainted attribute value.   
     
     
         22 . The computer-implemented method of  claim 21 , wherein the image is a raw image, further comprising:
 processing the raw image to generate a first image, wherein the first image is a hue and saturation (HS) channel.   
     
     
         23 . The computer-implemented method of  claim 22 , further comprising:
 processing the first image and the mask to generate a masked image; and   wherein the masked image is inputted into the machine-learned inpainting model to generate the inpainted image.   
     
     
         24 . The computer-implemented method of  claim 22 , further comprising:
 determining a palette transform based a comparison of the first image and the inpainted image, wherein the palette transform is generated using a plurality voting technique.   
     
     
         25 . The computer-implemented method of  claim 24 , wherein the palette transform is determined based on a dilated mask, the dilated mask having an expanded region of interest associated with the first image, the expanded region of interest of the dilated mask being larger than the region of interest of the mask. 
     
     
         26 . The computer-implemented method of  claim 22 , wherein the raw image is a high-resolution image, and the first image is a low-resolution image. 
     
     
         27 . The computer-implemented method of  claim 21 , wherein the final image is a high-resolution image, and the inpainted image is low-resolution image. 
     
     
         28 . The computer-implemented method of  claim 21 , wherein the one or more chromaticity channels comprise hue and saturation (HS) channels. 
     
     
         29 . The computer-implemented method of  claim 21 , wherein the machine-learned inpainting model is trained using hue and saturation (HS) training data. 
     
     
         30 . The computer-implemented method of  claim 21 , wherein the distractor object blends into a background of the final image. 
     
     
         31 . A computing system, comprising:
 one or more processors; and   one or more non-transitory computer-readable media that collectively store:   a machine-learned inpainting model, wherein the machine-learned inpainting model is configured to generate an inpainted image using an image; and   instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising:
 accessing the image and a mask, the image being in a red-green-blue (RGB) color space, wherein the mask indicates a region of interest associated with the image, and wherein a distractor object is inside the region of interest and has a first pixel with a first RGB value; 
 processing, using the machine-learned inpainting model, the image and the mask to generate an inpainted image, wherein the first pixel of the distractor object has a first inpainted attribute value in one or more chromaticity channels; 
 modifying, using a voting technique, the first pixel of the distractor object to a second inpainted attribute value in the one or more chromaticity channels, the second inpainted attribute value being different than the first inpainted attribute value; and 
 processing the image to generate a final image in the RGB color space, wherein the first pixel of the distractor object has a second RGB value that is different than the first RGB value, the second RGB value being based on the second inpainted attribute value. 
   
     
     
         32 . The computer system of  claim 31 , wherein the image is a raw image, the operation further comprising:
 processing the raw image to generate a first image, wherein the first image is a hue and saturation (HS) channel.   
     
     
         33 . The computer system of  claim 32 , the operation further comprising:
 processing the first image and the mask to generate a masked image; and   wherein the masked image is inputted into the machine-learned inpainting model to generate the inpainted image.   
     
     
         34 . The computer system of  claim 32 , the operation further comprising:
 determining a palette transform based a comparison of the first image and the inpainted image, wherein the palette transform is generated using a plurality voting technique.   
     
     
         35 . The computer system of  claim 34 , wherein the palette transform is determined based on a dilated mask, the dilated mask having an expanded region of interest associated with the first image, the expanded region of interest of the dilated mask being larger than the region of interest of the mask. 
     
     
         36 . The computer system of  claim 32 , wherein the raw image is a high-resolution image, and the first image is a low-resolution image. 
     
     
         37 . The computer system of  claim 31 , wherein the final image is a high-resolution image, and the inpainted image is low-resolution image. 
     
     
         38 . The computer system of  claim 31 , wherein the one or more chromaticity channels comprise hue and saturation (HS) channels. 
     
     
         39 . The computer system of  claim 31 , wherein the machine-learned inpainting model is trained using hue and saturation (HS) training data. 
     
     
         40 . One or more non-transitory, computer readable media storing instructions that are executable by one or more processors to cause a computing system to perform operations, the operations comprising:
 accessing an image and a mask, the image being in a red-green-blue (RGB) color space, wherein the mask indicates a region of interest associated with the image, and wherein a distractor object is inside the region of interest and has a first pixel with a first RGB value;   processing, using a machine-learned inpainting model, the image and the mask to generate an inpainted image, wherein the first pixel of the distractor object has a first inpainted attribute value in one or more chromaticity channels;   modifying, using a voting technique, the first pixel of the distractor object to a second inpainted attribute value in the one or more chromaticity channels, the second inpainted attribute value being different than the first inpainted attribute value; and   processing the image to generate a final image in the RGB color space, wherein the first pixel of the distractor object has a second RGB value that is different than the first RGB value, the second RGB value being based on the second inpainted attribute value.