Information processing apparatus and processing method for image data

1. An information processing apparatus including a processor and a memory and being configured to convert image data that has been input into saved data to save in a storage unit and reproduce the image data from the saved data, the information processing apparatus comprising:

an encoder unit configured to convert the image data into the saved data; and

a decoder unit configured to reproduce the saved data as the image data,

wherein the encoder unit includes:

a recognition unit configured to generate class tag information from the image data that has been input;

a segmentation unit configured to generate region information that distinguishes a recognition target region and a background region from the image data that has been input;

a region separation unit configured to generate a background image according to the background region from the image data that has been input based on the region information;

a feature extraction unit configured to generate a feature vector from the image data that has been input;

a reconstruction unit configured to generate a reconstructed image from the class tag information and the feature vector;

a region separation unit configured to generate a recognition target image and a background image from the image data that has been input and the region information; and

a differential extraction unit configured to generate a recognition differential image from the recognition target image and the reconstructed image,

the encoder unit outputs the saved data including the class tag information, the feature vector, the recognition differential image, and the background image, and

wherein the decoder unit includes:

a reconstruction unit configured to read the saved data and generate a reconstructed image from the class tag information and the feature vector; and

a merging unit configured to merge the recognition differential image, the background image, and the reconstructed image read from the saved data and reproduce the image data.

2. The information processing apparatus according to claim 1 , wherein

the saved data including the class tag information, the region information, and the background image is stored in a storage device.

3. The information processing apparatus according to claim 1 , wherein the storage unit is configured to store the saved data including the class tag information, the region information, and the background image generated by the encoder unit.

4. The information processing apparatus according to claim 1 , wherein

the recognition unit, the segmentation unit, and the reconstruction unit include neural networks.

5. The information processing apparatus according to claim 4 , wherein

in the neural networks of the recognition unit, the segmentation unit, and the reconstruction unit, weights for the neural networks are set through learning by inverse error propagation using teacher data, and

the teacher data includes a pair of preset learning image data and learning class tag information.

6. The information processing apparatus according to claim 4 , wherein

the recognition unit generates a class tag including a feature quantity of the image data from the image data that has been input.

7. The information processing apparatus according to claim 1 , further comprising a differential generation unit configured to generate a differential background image from the background image output by the region separation unit and the background image that has been output last time by the region separation unit.

8. The information processing apparatus according to claim 1 , wherein the storage unit is configured to store the saved data including the class tag information, the feature vector, the recognition differential image, and the background image generated by the encoder unit.

9. The information processing apparatus according to claim 1 , wherein

the recognition unit, the feature extraction unit, the segmentation unit, and the reconstruction unit include neural networks.

10. The information processing apparatus according to claim 9 , wherein

in the neural networks of the recognition unit, the feature extraction unit, the segmentation unit, and the reconstruction unit, weights for the neural networks are set through learning by inverse error propagation using teacher data, and

the teacher data includes a pair of preset learning image data and learning class tag information.

11. The information processing apparatus according to claim 9 , wherein

the recognition unit generates a class tag including a feature quantity of the image data from the image data that has been input.

12. The information processing apparatus according to claim 1 , wherein

the reconstruction unit of the encoder unit is the same as the reconstruction unit of the decoder unit.

13. A processing method for image data wherein a computer including a processor and a memory converts image data that has been input into saved data to save in a storage unit and reproduces the image data from the saved data, the processing method comprising:

generating class tag information from the image data that has been input;

generating region information that distinguishes a recognition target region and a background region from the image data that has been input;

generating a background image excluding the recognition target region from the image data that has been input based on the region information;

generating a feature vector from the image data that has been input;

generating a reconstructed image from the class tag information and the feature vector;

generating a recognition target image and a background image from the image data that has been input and the region information; and

generating a recognition differential image from the recognition target image and the reconstructed image,

wherein the computer outputs the saved data including the class tag information, the feature vector, the recognition differential image, and the background image,

the processing method further comprising:

reading the saved data and generating a reconstructed image from the class tag information and the feature vector; and

merging the recognition differential image, the background image, and the reconstructed image read from the saved data and reproduce the image data.

14. The processing method for image data according to claim 13 , further comprising:

storing the saved data including the class tag information, the region information, and the background image in a storage device.