Image magnification device and image magnification method

Abstract

A high-definition magnified image is obtained by simple computation. A magnified image is obtained by obtaining a deteriorated image on the basis of an input original image, determining enhancement filters used for individual small areas of the original image using the deteriorated image at least, performing a filtering process by applying the filters thus determined to the individual small areas to achieve enhancement, and interpolating the image for magnification.

Description

BACKGROUND OF THE INVENTION [0001] The present invention relates to a technique of magnifying images, which is applied, for example, to devices like mobile phones and PDAs (personal digital assistants) and to general software for displaying images. [0002] Various methods for magnification of images have been proposed. Most common methods apply filtering to an image after (or while) up-sampling the image. Bilinear interpolation and bicubic interpolation are the most famous examples of such methods. [0003] However, these methods tend to produce blurred magnified images. This is because high-frequency components are lost during image magnification. [0004] To solve the problem, Non-Patent document 1 describes a method in which, after magnification of an image, an enhancement / estimation process is performed for lost high-frequency components to realize high-definition image magnification. Also, Non-Patent document 2 describes a method in which signal pattern after magnification is estima...

AI Technical Summary

Benefits of technology

[0014] Accordingly, an object of the present invention is to provide a technique capable of obtaining a high-definition magnified image by simple computation.

Problems solved by technology

However, the conventional techniques cited above have serious problems.

Also, since uniform enhancement is applied to the entire image, the occurrence of noise due to the enhancement of high-frequency components is unavoidable.

Also, in the technique of Non-Patent document 2, the validity of associating an image signal before magnification and image pattern after magnification is questionable; a learned magnified image signal does not always correspond to the signal before magnification.

Furthermore, the technique of Non-Patent document 3 more greatly depends on the quality of learned data than the invention of Non-Patent document 2, while it is difficult to obtain general learned data.

Moreover, the solution search is extremely time-consuming and not practical.

Images