Systems and methods for image modification and image based content capture and extraction in neural networks
Systems and methods for image modification to increase contrast between text and non-text pixels within the image. In one embodiment, an original document image is scaled to a predetermined size for processing by a convolutional neural network. The convolutional neural network identifies a probability that each pixel in the scaled is text and generates a heat map of these probabilities. The heat map is then scaled back to the size of the original document image, and the probabilities in the heat map are used to adjust the intensities of the text and non-text pixels. For positive text, intensities of text pixels are reduced and intensities of non-text pixels are increased in order to increase the contrast of the text against the background of the image. Optical character recognition may then be performed on the contrast-adjusted image.
1. An image modification system comprising:
at least one processor; and
at least one non-transitory computer readable medium storing instructions that are executable by the at least one processor to cause the at least one processor to perform:
implementing a fully convolutional neural network;
receiving an input image from an input/output device communicatively connected to the image recognition system over a network;
if an original size of the input image does not match a predetermined image size which the convolutional neural network is configured to process, scaling the input image to the predetermined image size;
performing, using the fully convolutional neural network, a set of convolutional operations on the input image so as to produce at least a first heat map, wherein the first heat map indicates, for each pixel in the input image, a probability that the pixel forms a portion of a text character in the input image;
if the predetermined image size which the convolutional neural network is configured to process does not match the original size of the input image, scaling the first heat map to the original size of the input image;
modifying intensities of at least a subset of the pixels in the input image according to the first heat map to produce a modified image, wherein the intensities of a set of text pixels are adjusted in a first direction and a set of non-text pixels neighboring the set of text pixels are adjusted in a second direction opposite the first direction, thereby increasing a contrast between the set of text pixels and the set of neighboring non-text pixels, and wherein respective colors of the modified pixels remain the same before and after modifying the intensities of the modified pixels; and
storing the modified image in one or more memories coupled to the at least one processor.
2. The image modification system of claim 1 , wherein for each pixel in the input image, a magnitude of an intensity adjustment for the pixel is determined in dependence upon an initial intensity of the pixel in the input image and a probability value in the first heat map corresponding to the pixel.
3. The image modification system of claim 1 , wherein for each pixel in the input image, the pixel has a corresponding RGB (Red, Green, Blue) representation, wherein modifying the intensity of the pixel comprises converting the RGB representation of the pixel to an HSL (Hue, Saturation, Lightness) representation of the pixel, adjusting a lightness value of the HSL representation of the pixel, and converting the adjusted HSL representation of the pixel to an adjusted RGB representation of the pixel.
4. The image modification system of claim 1 , wherein for at least a portion of the pixels in the input image, the corresponding modified intensities of the pixels are less than a maximum intensity and greater than a minimum intensity.
5. The image modification system of claim 1 , wherein modifying the intensities of the pixels in the input image comprises:
for text pixels representing positive text characters, reducing the corresponding intensities of the pixels;
for text pixels representing inverted text characters, increasing the corresponding intensities of the pixels;
for non-text pixels neighboring positive text characters, increasing the corresponding intensities of the pixels; and
for non-text pixels neighboring inverted text characters, reducing the corresponding intensities of the pixels.
6. The image modification system of claim 1 , wherein the instructions are further executable by the at least one processor to cause the at least one processor to: retrieve the modified image from the one or more memories; perform optical character recognition on the modified image and thereby extract one or more text characters from the modified image; and
store the one or more extracted text characters in the one or more memories.
7. A method for modifying images, the method comprising:
receiving an input image from an input/output device communicatively connected to an image recognition system;
if an original size of the input image does not match a predetermined image size which the image recognition system is configured to process, scaling the input image to the predetermined image size;
generating, using the image recognition system, at least a first heat map, wherein the first heat map indicates, for each pixel in the input image, a probability that the pixel forms a portion of a text character in the input image;
if the predetermined image size which the image recognition system is configured to process does not match the original size of the input image, scaling the first heat map to the original size of the input image;
modifying intensities of at least a subset of the pixels in the input image according to the first heat map to produce a modified image, wherein the intensities of a set of text pixels are adjusted in a first direction and a set of non-text pixels neighboring the set of text pixels are adjusted in a second direction opposite the first direction, thereby increasing a contrast between the set of text pixels and the set of neighboring non-text pixels, and wherein respective colors of the modified pixels remain the same before and after modifying the intensities of the modified pixels; and
storing the modified image in one or more memories coupled to the image recognition system.
8. The method of claim 7 , wherein for each pixel in the input image, a magnitude of an intensity adjustment for the pixel is determined in dependence upon an initial intensity of the pixel in the input image and a probability value in the first heat map corresponding to the pixel.
9. The method of claim 7 , wherein for each pixel in the input image, the pixel has a corresponding RGB (Red, Green, Blue) representation, wherein modifying the intensity of the pixel comprises converting the RGB representation of the pixel to an HSL (Hue, Saturation, Lightness) representation of the pixel, adjusting a lightness value of the HSL representation of the pixel, and converting the adjusted HSL representation of the pixel to an adjusted RGB representation of the pixel.
10. The method of claim 7 , wherein for at least a portion of the pixels in the input image, the corresponding modified intensities of the pixels are less than a maximum intensity and greater than a minimum intensity.
11. The method of claim 7 , wherein modifying the intensities of the pixels in the input image comprises:
for text pixels representing positive text characters, reducing the corresponding intensities of the pixels;
for text pixels representing inverted text characters, increasing the corresponding intensities of the pixels;
for non-text pixels neighboring positive text characters, increasing the corresponding intensities of the pixels; and
for non-text pixels neighboring inverted text characters, reducing the corresponding intensities of the pixels.
12. The method of claim 7 , further comprising: retrieving the modified image from the one or more memories; performing optical character recognition on the modified image and thereby extracting one or more text characters from the modified image; and storing the one or more extracted text characters in the one or more memories.
13. A computer program product comprising a non-transitory computer-readable medium storing instructions executable by one or more processors to cause the one or more processors to perform:
receiving an input image from an input/output device communicatively connected to an image recognition system;
if an original size of the input image does not match a predetermined image size which the image recognition system is configured to process, scaling the input image to the predetermined image size;
generating, using the image recognition system, at least a first heat map, wherein the first heat map indicates, for each pixel in the input image, a probability that the pixel forms a portion of a text character in the input image;
if the predetermined image size which the image recognition system is configured to process does not match the original size of the input image, scaling the first heat map to the original size of the input image;
modifying intensities of at least a subset of the pixels in the input image according to the first heat map to produce a modified image, wherein the intensities of a set of text pixels are adjusted in a first direction and a set of non-text pixels neighboring the set of text pixels are adjusted in a second direction opposite the first direction, thereby increasing a contrast between the set of text pixels and the set of neighboring non-text pixels, and wherein respective colors of the modified pixels remain the same before and after modifying the intensities of the modified pixels; and
storing the modified image in one or more memories coupled to the image recognition system.
14. The computer program product of claim 13 , wherein for each pixel in the input image, a magnitude of an intensity adjustment for the pixel is determined in dependence upon an initial intensity of the pixel in the input image and a probability value in the first heat map corresponding to the pixel.
15. The computer program product of claim 13 , wherein for each pixel in the input image, the pixel has a corresponding RGB (Red, Green, Blue) representation, wherein modifying the intensity of the pixel comprises converting the RGB representation of the pixel to an HSL (Hue, Saturation, Lightness) representation of the pixel, adjusting a lightness value of the HSL representation of the pixel, and converting the adjusted HSL representation of the pixel to an adjusted RGB representation of the pixel.
16. The computer program product of claim 13 , wherein for at least a portion of the pixels in the input image, the corresponding modified intensities of the pixels are less than a maximum intensity and greater than a minimum intensity.
17. The computer program product of claim 13 , wherein modifying the intensities of the pixels in the input image comprises:
for text pixels representing positive text characters, reducing the corresponding intensities of the pixels;
for text pixels representing inverted text characters, increasing the corresponding intensities of the pixels;
for non-text pixels neighboring positive text characters, increasing the corresponding intensities of the pixels; and
for non-text pixels neighboring inverted text characters, reducing the corresponding intensities of the pixels.
18. The computer program product of claim 13 , further comprising: retrieving the modified image from the one or more memories; performing optical character recognition on the modified image and thereby extracting one or more text characters from the modified image; and storing the one or more extracted text characters in the one or more memories.