Image processing device, image recognition device, image processing program, and image recognition program

Abstract

The present invention reduces calculation costs while raising detection accuracy. An image processing device 8 has a function for plotting a luminance gradient co-occurrence pair of an image on a feature plane 15 and applying an EM algorithm thereto to form a GMM. The image processing device 8 learns a pedestrian image and creates a GMM of the pedestrian image, subsequently learns a background image and creates a GMM of the background image, and furthermore calculates a difference between the two and generates a GMM for relearning based on the calculation. The image processing device 8 plots asample that conforms to the GMM for relearning on the feature plane 15 by applying an inverse function theorem. The image processing device 8 finally forms a GMM that represents the distribution of samples at a designated mixed number and thereby forms a standard GMM that serves as a standard for image recognition. When this mixed number is set to less than a mixed number designated earlier, thedimensions with which an image is analyzed are reduced, making it possible to reduce calculation costs.

Description

technical field [0001] The present invention relates to an image processing device, an image recognition device, an image processing program, and an image recognition program, and relates to, for example, image processing using brightness gradients of pixels. Background technique [0002] As image recognition technology, there is a technique of extracting the distribution of gradient directions of brightness from an image as an image feature quantity, and comparing the distribution of gradient directions of brightness with the distribution of gradient directions of brightness of images learned in advance to identify objects. [0003] For example, the "image recognition device" of Patent Document 1 extracts feature points and their feature quantities from a standard image in advance. Also, feature points and their feature quantities are calculated from the input image, and feature points similar to feature quantities in the standard image are set as corresponding points. [...

AI Technical Summary

Problems solved by technology

[0007] However, in order to improve the accuracy, it is necessary to reduce the interval of quantization, but if the interval of quantization is reduced, there is a problem that the feature value becomes larger and the processing time increases.

Images