Image segmentation method and device

1. An image segmentation method, which performed in a computing device having one or more processors and a memory for storing one or more programs executed by the one or more processors, the method comprising the steps of:

(a) receiving an input image;

(b) generating a first-generation image segment set by dividing the input image in an overlapped manner;

(c) generating a second or higher-generation image segment set from the first-generation image segment set, wherein a subsequent-generation image segment set is generated by dividing in an overlapped manner at least one image segment that satisfies a preset selection condition among a plurality of image segments comprised in the previous-generation image segment set, wherein the preset selection condition is related to pixel similarity in an image; and

(d) generating a training image set including the input image and at least a part of image segments included in the first-generation image segment set and the subsequent-generation image segment set,

wherein the step (c) comprises:

calculating the pixel similarity in an image for each of the plurality of image segments comprised in the previous-generation image segment set by comparing each pixel in the image with other pixels in the same image based on at least one of brightness, color value, frequency and gradient of pixels; and

generating the subsequent-generation image segment set by dividing in an overlapped manner each of image segments that have the pixel similarity smaller than or equal to a preset reference value among the plurality of image segments comprised in the previous-generation image segment set.

2. The image segmentation method of claim 1 , wherein each of a plurality of image segments included in a j th -generation image segment set comprises a region that overlaps one or more other image segments included in the j th -generation image segment set, where j of the j th -generation image segment set is an integer greater than or equal to 1.

3. The image segmentation method of claim 1 , wherein the step (c) terminates generation of the subsequent-generation image segment set when there is no image segment whose pixel similarity is smaller than or equal to the reference value.

4. The image segmentation method of claim 1 , wherein the step (c) terminates generation of the subsequent-generation image segment set when a preset segmentation termination condition is satisfied.

5. The image segmentation method of claim 4 , wherein the step (c) determines whether the preset segmentation termination condition is satisfied based on a size of each of the image segments comprised in the previous-generation image segment set and a preset reference value.

6. The image segmentation method of claim 4 , wherein the step (c) determines whether the preset segmentation termination condition is satisfied based on a total number of image segments comprised in image segment sets generated so far and a preset reference value.

7. The image segmentation method of claim 1 , wherein the step (d) comprises:

reducing the input image and at least a part of the image segments included in each of the first to n th -generation image segment sets to a preset size, where n of the n th -generation image segment is an integer greater than or equal to 2.

8. The image segmentation method of claim 7 , further comprising:

(e) training a deep learning-based image classifier using the reduced images.

9. An image segmentation device comprising:

one or more processors;

a memory; and

one or more programs stored in the memory, the one or more programs configured to be executed by the one or more processors, the one or more programs comprising commands for performing the steps of:

(a) receiving an input image,

(b) generating a first-generation image segment set by dividing the input image in an overlapped manner,

(c) generating a second or higher-generation image segment set from the first-generation image segment set, wherein a subsequent-generation image segment set is generated by dividing in an overlapped manner at least one image segment that satisfies a preset selection condition among a plurality of image segments comprised in the previous-generation image segment set, wherein the preset selection condition is related to pixel similarity in an image, and

(d) generating a training image set including the input image and at least a part of image segments included in the first-generation image segment set and the subsequent-generation image segment set,

wherein the step (c) comprises:

calculating the pixel similarity in an image for each of the plurality of image segments comprised in the previous-generation image segment set by comparing each pixel in the image with other pixels in the same image based on at least one of brightness, color value, frequency and gradient of pixels; and

generating the subsequent-generation image segment set by dividing in an overlapped manner each of image segments that have the pixel similarity smaller than or equal to a preset reference value among the plurality of image segments comprised in the previous-generation image segment set.

10. The image segmentation device of claim 9 , wherein each of a plurality of image segments included in a j th -generation image segment set includes a region that overlaps one or more other image segments included in the j th -generation image segment set, where j of the j th -generation image segment set is an integer greater than or equal to 1.

11. The image segmentation device of claim 9 , wherein the step (c) terminates generation of the subsequent-generation image segment set when there is no image segment whose pixel similarity is smaller than or equal to the reference value.

12. The image segmentation device of claim 9 , wherein the step (c) terminates generation of the subsequent-generation image segment set when a preset segmentation termination condition is satisfied.

13. The image segmentation device of claim 12 , wherein the step (c) determines whether the preset segmentation termination condition is satisfied based on a size of each of the image segments comprised in the previous-generation image segment set and a preset reference value.

14. The image segmentation device of claim 12 , wherein the step (c) determines whether the preset segmentation termination condition is satisfied based on a total number of image segments comprised in image segment sets generated so far and a preset reference value.

15. The image segmentation device of claim 9 , wherein the step (d) comprises reducing the input image and a least a part of the image segments comprised in each of the first to n th -generation image segment sets to a preset size, where n of the n th -generation image segment is an integer greater than or equal to 2.

16. The image segmentation device of claim 15 , wherein the one or more programs further comprise a command for performing training a deep learning-based image classifier using the reduced images.