Image processing device, image processing method, computer program product
According to one embodiment, an image processing device includes a storage and an image processor. The storage stores therein an input image. The image processor segments the input image into a plurality of regions by using a first convolutional neural network (CNN), generates a first image by converting pixel values of pixels in a first region included in the regions into a first value, and performs image processing on the first image by using a second CNN to generate a second image.
1. An image processing device comprising:
a storage configured to store therein an input image; and
an image processor configured to:
use a first convolutional network (CNN) to segment the input image into a plurality of regions;
convert pixel values of pixels in a first region included in the regions into a same first value to generate a first image; and
use a second CNN to interpolate pixel values in the first region and reduce noise in regions other than the first region to generate a second image.
2. The image processing device according to claim 1 , wherein
the input image is a distance image including pixels having a value indicating a distance, and
the first region is a region estimated to be noisy by the first CNN.
3. The image processing device according to claim 1 , wherein the image processor is configured to:
generate a plurality of first images,
use any one of a plurality of second CNNs on each of the first images to generate a plurality of second images, and
combine the second images to generate a third image.
4. The image processing device according to claim 3 , wherein the image processor is configured to use a weighted average of pixel values of respective pixels in the second images as a pixel value of a corresponding pixel in the third image.
5. An image processing method comprising:
using a first convolutional network (CNN) to segment an input image into a plurality of regions;
converting pixel values of pixels in a first region included in the regions into a same first value to generate a first image; and
using a second CNN to interpolate pixel values in the first region and reduce noise in regions other than the first region to generate a second image.
6. A computer program product having a non-transitory computer readable medium including programmed instructions, wherein the instructions, when executed by a computer, cause the computer to perform:
using a first convolutional network (CNN) to segment an input image into a plurality of regions;
converting pixel values of pixels in a first region included in the regions into a same first value to generate a first image; and
using a second CNN to interpolate pixel values in the first region and reduce noise in regions other than the first region to generate a second image.