Color conversion between color spaces using reduced dimension embeddings
Exemplary embodiments may provide an approach to converting multidimensional color data for an image encoded in a first color space into an intermediate form that is a single dimensional value. The exemplary embodiments may then decode the intermediate form value to produce an encoding of the color data that is encoded in a second color space that differs from the first color space. In this manner, the data for the image may be efficiently converted from an encoding in the first color space into an encoding in the second color space.
1. A method, comprising:
accessing training data comprising an image associated with a first color space, the image comprising a first plurality of pixels;
training an encoder of a neural network executing on a processor to convert each respective pixel of the first plurality of pixels into a respective single-dimensional color value; and
training a decoder of the neural network to convert each respective single dimensional color value into a respective pixel of a second plurality of pixels, wherein the second plurality of pixels are in a second color space that is different than the first color space.
2. The method of claim 1 , wherein the decoder is one of a plurality of decoders of the neural network, the method further comprising:
training a second decoder of the plurality of decoders to convert each respective single dimensional color value into a respective pixel of a third plurality of pixels, wherein the third plurality of pixels are in a third color space that is different than the first color space and the second color space.
3. The method of claim 1 , wherein the encoder is one of a plurality of encoders, wherein the decoder is one of a plurality of decoders, the method further comprising:
selecting the encoder based on the first color space; and
selecting the decoder based on the second color space.
4. The method of claim 1 , further comprising:
encoding, by the trained encoder, each of a first plurality of pixels of another image into a respective single-dimensional color value, wherein the first plurality of pixels are based on the first color space; and
converting, by the trained decoder, each respective single dimensional color value of the another image into a respective pixel of a second plurality of pixels in the second color space.
5. The method of claim 1 , wherein the second plurality of pixels in the second color space are compressed relative to the first plurality of pixels in the first color space.
6. The method of claim 1 , wherein the single-dimensional color values are in a latent space of the neural network.
7. The method of claim 1 , wherein the image is a three-dimensional image, wherein a plurality of voxels of the three-dimensional image include the first plurality of pixels, wherein training the encoder comprises encoding the plurality of voxels into the respective single-dimensional color values.
8. The method of claim 1 , wherein the first color space or the second color space is one of a RGB color space, an LAB color space, an HSV color space, a CMYK color space, a YUV color space, a HSL color space, an ICtCp color space or a CIE color space.
9. The method of claim 1 , wherein the training data indicates a correct conversion of the single dimensional color values into the second color space, wherein the training of the encoder and the decoder is based on a comparison of the correct conversion to the second plurality of pixels.
10. A non-transitory computer-readable storage medium, the computer-readable storage medium including instructions that when executed by a processor, cause the processor to:
access training data comprising an image associated with a first color space, the image comprising a first plurality of pixels;
train an encoder of a neural network to convert each respective pixel of the first plurality of pixels into a respective single-dimensional color value; and
train a decoder of the neural network to convert each respective single dimensional color value into a respective pixel of a second plurality of pixels, wherein the second plurality of pixels are in a second color space that is different than the first color space.
11. The computer-readable storage medium of claim 10 , wherein the decoder is one of a plurality of decoders of the neural network, wherein the instructions further cause the processor to:
train a second decoder of the plurality of decoders to convert each respective single dimensional color value into a respective pixel of a third plurality of pixels, wherein the third plurality of pixels are in a third color space that is different than the first color space and the second color space.
12. The computer-readable storage medium of claim 10 , wherein the encoder is one of a plurality of encoders, wherein the decoder is one of a plurality of decoders, wherein the instructions further cause the processor to:
select the encoder based on the first color space; and
select the decoder based on the second color space.
13. The computer-readable storage medium of claim 10 , wherein the second plurality of pixels in the second color space are compressed relative to the first plurality of pixels in the first color space.
14. The computer-readable storage medium of claim 10 , wherein the single-dimensional color values are in a latent space of the neural network.
15. The computer-readable storage medium of claim 10 , wherein the image is a three-dimensional image, wherein a plurality of voxels of the three-dimensional image include the first plurality of pixels, wherein training the encoder comprises encode the plurality of voxels into the respective single-dimensional color values.
16. A computing apparatus comprising:
a processor; and
a memory storing instructions that, when executed by the processor, cause the processor to:
access training data comprising an image associated with a first color space, the image comprising a first plurality of pixels;
train an encoder of a neural network to convert each respective pixel of the first plurality of pixels into a respective single-dimensional color value; and
train a decoder of the neural network to convert each respective single dimensional color value into a respective pixel of a second plurality of pixels, wherein the second plurality of pixels are in a second color space that is different than the first color space.
17. The computing apparatus of claim 16 , wherein the decoder is one of a plurality of decoders of the neural network, wherein the instructions further cause the processor to:
train a second decoder of the plurality of decoders to convert each respective single dimensional color value into a respective pixel of a third plurality of pixels, wherein the third plurality of pixels are in a third color space that is different than the first color space and the second color space.
18. The computing apparatus of claim 16 , wherein the second plurality of pixels in the second color space are compressed relative to the first plurality of pixels in the first color space.
19. The computing apparatus of claim 16 , wherein the single-dimensional color values are in a latent space of the neural network.
20. The computing apparatus of claim 16 , wherein the training data indicates a correct conversion of the single dimensional color values into the second color space, wherein the training of the encoder and the decoder is based on a comparison of the correct conversion to the second plurality of pixels.