Image processing method and image processing device

Abstract

The present invention provides an image processing method and device realizing a high versatility and a learning accuracy and having a designability of high degree of freedom. A medical image prepared as a teacher input image is subjected to different image processings during learning to create teacher feature images, and a teacher output image is created from the medical image. The pixel values of the pixels at the same position of the teacher input image including the teacher feature images is inputted into a discriminator (20). Learning of the discriminator (20) is carried out so that the difference between the output value from the discriminator (20) and the pixel value of the pixel at the pixel position in the teacher output image can be reduced. In emphasization, a feature image is created from an image to be processed. The pixel value of the pixel at the same position in these images is inputted into the discriminator (20). An emphasis image in which the output value outputted from the discriminator (20) is set as the pixel value of the pixel at the pixel position is outputted.

Description

technical field [0001] The present invention relates to an image processing method and an image processing device for outputting an output image in which a specific pattern is emphasized from an input image. Background technique [0002] Conventionally, there is a method of pattern recognition using a recognizer such as an artificial intelligence network (hereinafter abbreviated as ANN) or a support vector machine (support vector machine), which uses so-called teacher data for learning. Sampling data learns a classifier for a specific pattern having a characteristic shape, shape, color, density, size, and the like. [0003] A device has been developed that uses this method to perform pattern recognition on an image region predicted to be a shadow of a diseased part (referred to as an abnormal shadow) from a medical image obtained by inspection photography in the medical field, and Candidate regions detected as anomalous shadows. This device is called CAD (Computer Aided Di...

AI Technical Summary

Problems solved by technology

[0010] However, in the methods described in the above-mentioned Patent Documents 1 and 2, there are few analysis elements of the recognizer, and it is difficult to logically analyze how the output value obtained from the recognizer is affected by the input value and is output in a black-box manner.

Therefore, for practical use such as adoption as an algorithm of CAD, there may be restrictions on the degree of freedom of design.

Images