Learning apparatus and method, image processing apparatus and method, program, and recording medium

1. An image processing apparatus comprising:

a central processing unit operable to:

generate an estimated low-resolution image from a first high-resolution image output one frame before, wherein the estimated low-resolution image is generated by performing motion compensation on the first high-resolution image; and

predict a pixel value for a second high-resolution image to be output based on the estimated low-resolution image, an observed low-resolution image, first parameters stored in advance and second parameters stored in advance,

wherein the pixel value for the second high-resolution image is predicted based on summation of products of the first parameters and corresponding pixel values of the observed low-resolution image, and products of the second parameters and corresponding pixel values of the motion compensated first high-resolution image.

2. The image processing apparatus according to claim 1 , wherein the central processing unit is operable to read the first parameters and the second parameters stored in a learning database generated in advance by learning.

3. The image processing apparatus according to claim 1 , wherein the central processing unit is operable to:

compensate for a motion of the first high-resolution image;

restrict a band of the motion compensated first high-resolution image according to a camera model, thin out pixels of the restricted motion compensated first high-resolution image, and generate the estimated low-resolution image;

compute a difference of the observed low-resolution image that is an input low-resolution image and the estimated low-resolution image, and generate difference image data; and

interpolate pixels of the difference image data and thereby generate the second high-resolution image.

4. The image processing apparatus according to claim 1 , wherein the central processing unit is operable to:

add blur to the first high-resolution image; and

remove the blur from an up-sampled difference image,

wherein the difference image is generated from a down-sampled first high-resolution image and the observed low-resolution image.

5. The image processing apparatus according to claim 1 , wherein the central processing unit is operable to:

sort attention pixels for each class based on a feature amount, wherein the feature amount is a difference value between a pixel of the estimated low-resolution image and a corresponding pixel of a low-resolution student image; and

perform a predetermined operation using the first parameters and the second parameters corresponding to sorted classes stored in advance and predict the pixel value of the second high-resolution image.

6. The image processing apparatus according to claim 1 , wherein the central processing unit is operable to:

extract predetermined pixels of the observed low-resolution image that is an input low-resolution image and predetermined pixels of the estimated low-resolution image obtained with respect to the first high-resolution image as taps; and

perform a predetermined operation regarding the taps based on the first parameters and the second parameters stored in advance, and predict the pixel value of the second high-resolution image.

7. An image processing method comprising:

in a central processing unit:

generating an estimated low-resolution image from a first high-resolution image output one frame before, wherein the estimated low-resolution image is generated by performing motion compensation on the first high-resolution image; and

predicting a pixel value for a second high-resolution image to be output based on the estimated low-resolution image, an observed low-resolution image, first parameters stored in advance and second parameters stored in advance,

wherein the pixel value for the second high-resolution image is predicted based on summation of products of the first parameters and corresponding pixel values of the observed low-resolution image, and products of the second parameters and corresponding pixel values of the motion compensated first high-resolution image.

8. The image processing method of claim 7 , further comprises predicting the pixel value once with respect to an image frame to generate the second high-resolution image.

9. A non-transitory computer-readable storage medium having stored thereon, a computer program having at least one code section for image processing, the at least one code section being executable by a computer for causing the computer to perform steps comprising:

generating an estimated low-resolution image from a first high-resolution image output one frame before, wherein the estimated low-resolution image is generated by performing motion compensation on the first high-resolution image; and

predicting a pixel value for a second high-resolution image to be output based on the estimated low-resolution image, an observed low-resolution image, first parameters stored in advance and second parameters stored in advance,

wherein the pixel value for the second high-resolution image is predicted based on summation of products of the first parameters and corresponding pixel values of the observed low-resolution image, and products of the second parameters and corresponding pixel values of the motion compensated first high-resolution image.